A.8.20 Technological
Networks security
Structured attributes from ISO/IEC 27002:2022 — control type · CIA properties · cybersecurity concept · operational capability · security domain. What do these mean?
Mapped NIST 800-53 r5 controls (21)
Our reading (mostly llm_verified) of how this ISO control and each NIST 800-53 control relate. Not an ISO or NIST product.
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Why these map — rationale
- SC-7mostlycovers — Both controls establish boundary protection mechanisms such as firewalls, connection filtering, and device authentication to safeguard network traffic and connected systems.
- SC-8mostlycovers — A.8.20's broad mandate to protect network information and facilities from compromise directly accounts for the bulk of SC-8's transmission confidentiality/integrity requirement (via network protections like encryption and integrity controls), but leaves a residual slice of SC-8 uncovered for non-network transmission paths or mechanisms.
- AC-4partialcovers — A.8.20's broad network protection objective (preventing compromise via the network) addresses a slice of ac-4's information flow enforcement but leaves the bulk of the target's specific authorization-based flow control mechanisms and policy enforcement uncovered
Aligned NIST CSF 2.0 outcomes (21)
NIST CSF 2.0 outcomes this ISO control aligns with — our analysis (mostly llm_verified).
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Why these map — rationale
- PR.IR-01mostlycovers — The ISO control's emphasis on network segmentation, connection filtering, device authentication, and protocol hardening directly implements the CSF outcome of protecting networks and environments from unauthorized logical access.
- DE.CM-01domain mandateimplements — A.8.20's technical network-protection measures give operational effect to DE.CM-01's monitoring outcome within the shared network-security domain, but without either naming the other
Related OWASP ASVS 5.0 requirements (13)
Application-security verification requirements (OWASP ASVS 5.0) this ISO control aligns with; links open the ASVS chapter. Our analysis (mostly llm_verified) — many ISO controls have no ASVS counterpart.
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Related weaknesses / CWE (3269)
Weakness classes this ISO control helps prevent or mitigate — our analysis (mostly llm_verified).
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Why these map — rationale
- CWE-419mostlyblocks — A.8.20 directly mandates restricting/filtering network connections (firewalls), segregating channels, hardening devices, disabling vulnerable protocols, authenticating systems, and protecting channels over public/wireless networks, which stops exploitation of an unprotected primary admin channel in most cases; the bounded remainder is product-level channel protections inside an application that are outside network-scope.
- CWE-1125partialblocks — A.8.20's firewalls, filtering, authentication, and disabled protocols block exploitation of the network-reachable portion of the attack surface from untrusted networks, but the attack surface itself is not reduced and application-level or internally-reachable input/output points remain exploitable, leaving a nameable remainder.
- CWE-1188partialblocks — A.8.20's network restrictions, firewalls, and disabled protocols can block remote exploitation of insecure defaults that are network-reachable, but the insecure resource still remains exploitable by local users, insiders, or application-level defaults that network controls do not govern.
- CWE-1240partialblocks — Network-layer encryption, protocol disabling, and subnet isolation can block network-borne exploitation of risky cryptographic primitives, but the weak implementation remains breakable wherever its ciphertext or endpoint is reachable.
- CWE-1277partialblocks — A.8.20's network controls (firewalls, filtering, segmentation, isolation of subnetworks) can reduce the attack surface and block some exploitation paths to devices with unpatchable firmware, but the firmware remains vulnerable and reachable through legitimate network channels or physical access, so the weakness is still exploitable in many cases.
- CWE-1327partialblocks — A.8.20's network-level controls (firewalls, connection filtering per items i and j) can restrict which external IPs reach a service bound to 0.0.0.0, blocking exploitation from outside the permitted networks; however, the binding itself is unchanged, local/trusted-network actors can still reach the service, and the control does not address the misconfiguration that introduced the weakness.
- CWE-1393partialblocks — A.8.20 can restrict and filter network connections, segregate administration channels and isolate segments, which reduces remote exploitability of default passwords, but it does not stop exploitation by anyone already on a permitted network path or with local access.
- CWE-1428partialblocks — A.8.20's network-level encryption (VPN, IPsec per guidance item e) masks HTTP traffic in transit over the network, denying the attacker use of the cleartext during that portion of the path -- the same partial-transit reading the owner applied to A.8.20 vs CWE-523 -- but the product still uses HTTP, so the application-layer exploitation (cache poisoning, phishing, server spoofing) and any traffic outside the encrypted network segment remain exploitable.
- CWE-205partialblocks — A.8.20's firewall/filtering (item i), network access restriction (item j), and protocol disabling (item n) block external probing traffic that would trigger observable behavioral discrepancies, but the weakness remains in the product itself and is exploitable by any actor who can reach the product — the network control masks part of the exploitation surface without addressing the underlying behavioral leak.
- CWE-288partialblocks — A.8.20's network-level controls (firewalls per (i), network authentication per (h), connection filtering per (j)) can block exploitation of network-accessible alternate paths — e.g., a directly reachable web management URL or login screen — by restricting who reaches them, but cannot address local alternate paths (keyboard sequences, hardware registers), application-level alternate channels within the same service (API vs GUI), or design-level shadow resources, leaving a nameable remainder.
- CWE-290partialblocks — A.8.20's network-level controls (firewalls, ingress filtering, network authentication per items h/i/j/n) can block some spoofing attempts at the network layer, but the application-level authentication defect that trusts spoofable values remains present and exploitable through other vectors (DNS spoofing, Host header manipulation, on-path attacks within the trusted network).
- CWE-291partialblocks — A.8.20's network-level controls (firewalls, ingress filtering/anti-spoofing, restricting and authenticating system connections per items h, i, j) block a portion of IP-spoofing exploitation attempts against applications that rely on IP for authentication, but the application-level design flaw remains and is still exploitable via same-subnet attackers, source routing, or any spoofed traffic that traverses the network defenses.
- CWE-294partialblocks — A.8.20 item (e) mandates confidentiality controls on data passing over public, third-party and wireless networks, which prevents capture of authentication traffic over those segments and so blocks the replay attack there; the underlying design weakness (no replay protection in the protocol) remains, and capture on segments the clause does not reach (local networks, application-layer paths) is unaffected, mirroring the A.8.20 vs CWE-523 partial reading.
- CWE-306partialblocks — Network controls (firewalls, network authentication, filtering, segmentation) can block external or unauthorized network access to an unauthenticated critical function, but the missing authentication remains in the application and is still exploitable by anyone who reaches the service path — including internal users, anyone bypassing the firewall, or cases where the function is intentionally network-exposed.
- CWE-353partialblocks — A.8.20(e) requires safeguarding confidentiality AND integrity of data passing over networks, which can be implemented via network-layer integrity mechanisms (IPsec AH, MACsec, VPN with integrity) that mask the missing application-protocol checksum during transit; the weakness remains in the protocol itself and protection is bounded to network transit, leaving endpoint-side and intra-trust-boundary segments exposed, paralleling the A.8.20 -> CWE-523 `blocks` `partial` anchor.
- CWE-354partialblocks — A.8.20's network-level controls (encryption in transit per (e), firewalls per (i), disabling vulnerable protocols per (n)) can mask the vulnerability during a portion of transit by preventing in-flight modification or filtering unauthorized sources, but the weakness is an application-layer defect in the receiving product's checksum validation, which remains exploitable once data is decrypted at the network boundary or originates from within a trusted segment.
- CWE-419partialremoves — A.8.20 mandates network-level protections for primary channels — segregation of admin channels (l), authentication (h), hardening (k) — so where the unprotected channel is a network-layer issue, applying the clause removes the weakness; the nameable remainder is application/product-level unprotected channels (e.g., an unauthenticated web admin interface), which a network-security clause cannot redesign.
- CWE-420partialblocks — A.8.20's network filtering, segmentation, and disabling of vulnerable protocols can stop remote exploitation of some network-reachable alternate channels, but local, hardware, and application-level alternate channels are outside network security and remain exploitable.
- CWE-489partialblocks — A.8.20's firewalls, network authentication and segregation can block external network access to debug interfaces, denying the attacker use of the weakness over part of the path, but the debug code remains present and exploitable from within the network or via any interface the controls do not reach.
- CWE-494partialblocks — A.8.20's network controls (encryption of data in transit per (e), firewalls/filtering per (i), disabling vulnerable protocols per (n)) block some attack vectors for CWE-494 — in-transit modification and unauthorized access to update servers — but the root weakness (the product not verifying integrity of downloaded code) remains, so an attacker who bypasses network controls (e.g., compromises a legitimate server or tricks the client) can still exploit it.
- CWE-5partialblocks — A.8.20 item (e) requires controls to safeguard data passing over public, third-party, or wireless networks, which blocks exploitation of CWE-5's plaintext transmission over those network segments, but the clause does not mandate application-layer encryption (e.g., HTTPS) for internal or local network traffic, leaving a nameable remainder where the weakness can still be exploited.
- CWE-508partialblocks — Network controls (firewalls, filtering, isolation per items i, j, m) can block some exploitation paths of malicious code already on a system — particularly C2 communication, exfiltration, and lateral movement — but the code can still execute locally and cause harm, so only a portion of exploitation is blocked.
- CWE-509partialblocks — A.8.20's segmentation, connection filtering, and disabling of vulnerable protocols can block some network-based propagation of malicious code, but it does not inspect allowed application traffic or endpoint behavior, so viruses and worms spreading through permitted channels remain exploitable.
- CWE-523partialblocks — Requires network security controls that can protect credential transmission.
- CWE-614partialblocks — A.8.20's network-level encryption and traffic controls (e.g., forcing HTTPS via firewall rules on controlled networks) can mask the vulnerability during a portion of transit, but the weakness is in the application's cookie configuration and remains exploitable when traffic leaves the controlled network or when the application itself does not enforce HTTPS — the same scope reasoning the owner applied to A.8.20 vs CWE-523.
- CWE-693partialblocks — A.8.20 implements network-layer protection mechanisms (firewalls, network segmentation, transit encryption, connection filtering, device hardening) that, when well implemented, block exploitation of attacks relying on missing or insufficient network protection; but CWE-693 is a Pillar weakness covering all protection mechanism failures across the product, and the vast majority—application-layer input validation, output encoding, authentication, authorization, memory protection, cryptographic storage—are outside A.8.20's scope.
- CWE-757partialblocks — A.8.20 items (k) hardening of network devices and (n) disabling vulnerable network protocols can block some exploitation of algorithm downgrade by removing weak protocols from service or configuring devices to prefer strong algorithms, but the weakness lives in the protocol's negotiation logic—an application-layer design concern the clause does not reach—so most instances survive.
- CWE-770partialblocks — Network-level controls in A.8.20 (firewalls with rate limiting, connection filtering, disabling vulnerable protocols) can block some flooding-based resource exhaustion attacks (e.g., TCP/SIP/HTTP floods that exhaust connections or file descriptors), but the underlying application defect — the product allocating resources without limits — remains present and exploitable through single-request vectors (large image bands, oversized MIME parts, large length values) that network controls do not reach.
- CWE-778partialprevents — A.8.20 item (f) requires appropriate logging of actions affecting information security in the network context, which prevents CWE-778 instances related to network security events; however, CWE-778 is much broader (failed logins, authentication events, application-level events per the CVE examples), and A.8.20's logging mandate is scoped to network actions and lacks the event-content specification that A.8.15 carries, so only a minority of CWE-778 instances are prevented.
- CWE-918partialblocks — A.8.20's firewall and network filtering controls (items i, j, n) can block a meaningful portion of SSRF exploitation paths — particularly outbound connections to internal networks, cloud metadata endpoints, and disallowed protocols — but the application-layer URL validation defect remains and SSRF can still be exploited via external attacker-controlled destinations, DNS rebinding, and allowed internal services, leaving a nameable remainder.
- CWE-924partialblocks — A.8.20 item (e) mandates confidentiality and integrity of data passing over networks, and items (k), (l), (n) provide network-layer integrity mechanisms (hardening, segregated admin channels, disabling vulnerable protocols) that block exploitation during the network portion of transit — the same scope logic as the A.8.20 vs CWE-523 anchor. The product-level weakness remains: the receiving endpoint still does not check integrity itself, so if network-layer integrity is terminated at a gateway, bypassed on an alternate path, or absent on a local segment the clause does not reach, the message can still be modified in transit and the product will accept it.
- CWE-940partialblocks — Network-level controls (firewalls, network authentication, filtering) can restrict which systems can reach a service, blocking some exploitation paths for CWE-940 (e.g., restricting which hosts can send DNS updates), but the fundamental application-level source verification failure remains and many exploitation paths (Android intents, WebView injection) are entirely within the device and outside any network control's reach.
- CWE-941partialblocks — Network filtering, firewalling, isolation, and authentication can stop some unintended-destination communications, but A.8.20 does not ensure correct destination specification and leaves many SSRF, spoofing, and misrouting exploits possible.
Mitigated MITRE ATT&CK techniques (3586)
Adversary techniques (MITRE ATT&CK Enterprise) this ISO control helps mitigate; links open attack.mitre.org. Our analysis (mostly llm_verified).
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Why these map — rationale
- T1055mostlydetects — A.8.20 explicitly requires logging and monitoring (item f, cross-referencing 8.16) of actions affecting information security, which surfaces process injection in flight on monitored hosts and networks; mostly because the clause sets scope by classification/responsibilities rather than mandating universal deep instrumentation, leaving a bounded remainder for unmonitored processes or virtualized segments outside the defined monitoring perimeter.
- T1055mostlyresponds — A.8.20 item m (temporarily isolating critical subnetworks with drawbridges if the network is under attack) and the broader network monitoring/logging in f directly engage containment once process injection is underway on the network, bounding its spread via lateral movement or C2 while the remainder (in-process injection without observable network artifacts) is addressed by referenced controls such as 8.16.
- T1071.001mostlydetects — A.8.20 explicitly requires logging and monitoring of network actions (item f, cross-referenced to 8.16) plus detection-oriented items such as restricting/filtering connections (i), detecting device connections (j), and isolating subnetworks under attack (m); these surface anomalous or unauthorized web-protocol C2 traffic that blends with legitimate flows, though coverage is not exhaustive for every possible concealment technique inside common HTTP/S or WebSocket fields.
- T1071.002mostlydetects — A.8.20 explicitly requires logging and monitoring of network activity (item f, cross-referencing 8.16) plus detection-oriented measures such as restricting/filtering connections (i), detecting device connections (j), and isolating subnetworks under attack (m), all of which surface anomalous use of common file-transfer protocols like SMB/FTP that blend with legitimate traffic.
- T1071.003mostlydetects — A.8.20 explicitly requires logging and monitoring (item f, cross-referencing 8.16) of network actions plus detection of device connections (item j), network traffic filtering (item i), and protocol hardening (item n), which surfaces anomalous or concealed use of common mail protocols like SMTP/POP3/IMAP in T1071.003; the bounded remainder is traffic that perfectly mimics expected patterns without triggering the implemented monitoring scope.
- T1090.001mostlydetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16), restricting/filtering connections via firewalls (i), detecting device connections (j), and hardening/isolation of vulnerable protocols or attacked subnetworks (k,m,n), all of which surface internal proxy traffic that deviates from baselines or uses unexpected redirection.
- T1110.003mostlydetects — A.8.20 explicitly requires logging and monitoring of network actions (item f, cross-referencing 8.16) plus detection-oriented items (j: detecting device connections; m: isolating subnetworks under attack; n: disabling vulnerable protocols), which surface password-spraying attempts against the listed management ports/services on the network; the remainder is throttling or non-network vectors (e.g. cloud SSO) that fall outside network-scope monitoring.
- T1205.001mostlydetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16) plus detection-oriented items (j: detecting device connections; i: filtering with firewalls; m: isolating subnetworks under attack; n: disabling vulnerable protocols), all of which surface the anomalous port-knocking sequence on the monitored estate; the bounded remainder is custom raw-socket or libpcap implementations that produce no observable network artifact distinguishable from legitimate closed-port traffic.
- T1210mostlydetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16), hardening (k), restricting/filtering connections (i), and detecting device connections (j), all of which surface exploitation attempts or anomalous remote-service activity on the network; the remainder is post-exploitation behavior that succeeds without observable network artifacts.
- T1219mostlydetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16), hardening (k), restricting/filtering connections via firewalls (i), detecting device connections (j), and isolating subnetworks under attack (m); these surface post-compromise use of remote access tools on the network (e.g. anomalous sessions, new device connections, or C2 traffic patterns) across the great majority of the technique's network-visible behaviors, with only a bounded remainder (purely local non-networked use after installation) outside its scope.
- T1486mostlyresponds — A.8.20 item m (temporarily isolating critical subnetworks with drawbridges if the network is under attack) plus logging/monitoring (f) and coordination (g) directly enable containment of an in-progress ransomware encryption campaign once underway, bounding spread across the network while other items like hardening or protocol disabling sit outside the verb
- T1499.002mostlydetects — A.8.20 explicitly requires logging/monitoring of actions affecting network security (item f, cross-referencing 8.16), hardening (k), protocol disabling (n), and isolating subnetworks under attack (m), all of which surface service-exhaustion floods in flight on the organization's networks and devices; the bounded remainder is attacks against external third-party services outside the monitored estate.
- T1557mostlyprevents — A.8.20's mechanisms directly prevent several AiTM vectors — disabling vulnerable protocols (n) removes LLMNR/mDNS abuse, 802.1X-style authentication (h, j) blocks rogue device positioning, network segmentation limits broadcast-domain attacks, and hardening (k) reduces device compromise — but ARP spoofing without DAI, DNS cache poisoning at the resolver, physical access, and compromised infrastructure devices remain reachable vectors the clause does not fully close.
- T1557.001mostlyprevents — A.8.20 item (n) explicitly calls for disabling vulnerable network protocols, and LLMNR/NBT-NS/mDNS are the canonical vulnerable protocols this technique abuses — disabling them removes the traffic to poison. The remainder is that the control lists this as something to consider rather than universally mandating it, and the SMB relay half of the technique can occur independently of name resolution poisoning when an adversary is already in the authentication path.
- T1570mostlydetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16) plus network separation, authentication, connection filtering/firewalls, device connection detection, and protocol hardening, all of which surface lateral file-transfer activity over SMB/RDP/scp/etc. on the monitored estate; the bounded remainder is transfers that use already-allowed paths with no anomalous signature.
- T1601mostlyprevents — A.8.20's device hardening, connection restrictions, system/device authentication, administrative-plane segregation, and disabling of vulnerable protocols prevent most network-based modification of network device system images, though they cannot eliminate modification by already-privileged insiders, physical access, or zero-day exploitation.
- T1602.001mostlydetects — A.8.20 explicitly requires logging and monitoring of network actions (item f, cross-referencing 8.16) plus detection-oriented items such as j (detecting device connections), m (isolating subnetworks under attack), and n (disabling vulnerable protocols); these surface SNMP MIB queries on managed network devices, with the bounded remainder being stealthy or non-network-visible queries outside the monitored scope.
- T1602.001mostlyprevents — A.8.20's guidance items (k) hardening of network devices, (n) disabling vulnerable network protocols, and (i)/(j) filtering and restricting connections enable strong prevention of unauthorized SNMP MIB queries — disabling SNMP removes the attack surface, SNMPv3 with authentication blocks rogue queries, and boundary filtering blocks external access; the bounded remainder is SNMP legitimately retained for management and reachable from an internal foothold, where the technique can still run.
- T1602.002mostlyprevents — A.8.20 directly addresses the technique's primary vectors: hardening (k) and disabling vulnerable network protocols (n) target SMI and misconfigured SNMP, while filtering (i), authentication (h), and segregated admin channels (l) restrict who can reach management interfaces. The remainder is the adversary who already holds valid credentials or has compromised the device through another path — prevention cannot reach what authentication and filtering have already admitted.
- T1669mostlyprevents — A.8.20 explicitly addresses wireless networks in clause (e) and provides multiple mechanisms — strong authentication of systems on the network (h), restricting/filtering connections (i), detecting and authenticating equipment connections (j), hardening network devices (k), and disabling vulnerable protocols (n) — that prevent unauthorized Wi-Fi access when well implemented; the named remainder is the valid-accounts path the technique itself flags, where possession of legitimate credentials bypasses the wireless-layer controls.
- T1001partialdetects — A.8.20(f) explicitly mandates network logging and monitoring for detection, and T1001 manifests on the network — so the vantage point is correct — but the technique is purpose-built to look like legitimate traffic, and A.8.20 sets general monitoring scope rather than mandating C2-specific deep inspection, so only a slice of obfuscation (statistical anomalies, junk-data patterns, protocol misuse) is surfaced.
- T1001partialresponds — A.8.20's guidance includes explicit response-like mechanisms — temporarily isolating critical subnetworks under attack (m), filtering/restricting connections (i), and disabling vulnerable protocols (n) — that can act on T1001's obfuscated C2 channels once underway, but these are a slice of a primarily preventive control and the full incident response procedure belongs to A.5.26.
- T1001.001partialdetects — A.8.20's logging and monitoring (item f) and network filtering (item i) can surface anomalous traffic patterns and protocol irregularities that may indicate junk-data C2, but the technique is specifically designed to defeat trivial traffic analysis, and A.8.20 sets scope rather than mandating the deep protocol/DPI analysis needed to reliably distinguish junk-padded C2 from benign traffic.
- T1001.002partialresponds — A.8.20 provides the capability to temporarily isolate critical subnetworks during an attack, which contains the steganographic command and control traffic by severing the network connection, but it does not eradicate the underlying malware generating the traffic.
- T1001.003partialdetects — A.8.20's item (f) explicitly mandates logging and monitoring to enable detection of actions relevant to information security, and items (i), (j) and (k) provide filtering, connection detection and hardening that can surface some protocol anomalies; however, T1001.003 is specifically engineered to make C2 traffic blend with legitimate services (fake SSL/TLS handshakes, mimicked HTTP headers and certificates), and A.8.20 sets a network-security framework rather than mandating the deep protocol-validation capabilities (certificate chain inspection, JA3/JA3S fingerprinting, protocol-state analysis) needed to reliably distinguish impersonated traffic from genuine traffic, so what is caught is a slice rather than the bulk.
- T1001.003partialresponds — A.8.20 contains network-level response capabilities — temporarily isolating critical subnetworks (m), disabling vulnerable protocols (n), and filtering connections (i) — that can act on T1001.003's impersonated C2 traffic once detected, but it is not an incident response procedure and does not cover assessment, evidence collection, communication, or closure, so the response is a slice of the full verb.
- T1003.003partialdetects — A.8.20's monitoring scope (item f) covers network actions and could surface exfiltration of NTDS.dit or unusual connections to domain controllers, but the core of T1003.003 is host-based file access and tool execution (ntdsutil.exe, secretsdump.py, shadow copy creation) on the domain controller itself, which network security controls do not see.
- T1003.006partialdetects — A.8.20 item (f) requires logging and monitoring of network actions, which can surface DCSync replication traffic from non-domain-controller sources or anomalous replication patterns; however, DCSync uses legitimate DRSUAPI protocols designed to mimic normal DC-to-DC replication, and the clause's monitoring scope is network-level rather than AD-specific, so a compromised DC or well-crafted replication request would evade detection.
- T1003.006partialprevents — A.8.20's network filtering, system authentication, and segregation of administration channels can restrict which systems are permitted to initiate replication requests to a domain controller, narrowing the attack surface for DCSync; however, the technique uses legitimate DRSR replication protocols that any DC must accept, and the clause does not address credential/privilege protection or DC-specific hardening, so a compromised admin account on a permitted system can still execute it.
- T1003.006partialresponds — A.8.20 clause (m) — temporarily isolating critical subnetworks via drawbridges when the network is under attack — is a genuine containment action that can stop ongoing DCSync replication requests, and clause (i) filtering can be tightened to block replication traffic; however, this is a slice of the full response needed because the hashes are already exfiltrated, the adversary's foothold is not eradicated by network controls, and DCSync abuses legitimate DRSUAPI replication that is difficult to distinguish from normal traffic at the network layer.
- T1005partialdetects — A.8.20's network-level logging and monitoring (clause f) and network device authentication/hardening can detect T1005 against the Network Devices platform (e.g., unauthorized CLI access to read configs) and surface some network-side indicators of local collection, but the bulk of T1005 — local file system, configuration, database, and process memory reads on Windows, Linux, macOS and ESXi — happens below the network layer and is outside A.8.20's monitoring scope.
- T1008partialdetects — A.8.20's item (f) requires logging and monitoring of network actions, which can surface some fallback channels (unusual outbound flows, beaconing, DNS tunneling), but fallback channels are designed to blend with legitimate traffic and A.8.20's primary focus is network security controls (firewalls, segmentation, hardening) rather than deep traffic analysis, so detection is a slice rather than a bounded remainder of a working mechanism.
- T1008partialprevents — A.8.20 mandates network segmentation, traffic filtering, and disabling vulnerable protocols, which can block specific fallback channels, but adversaries can use allowed protocols or out-of-band channels that the network controls do not inspect.
- T1008partialresponds — A.8.20 includes explicit response mechanisms — (m) temporarily isolating critical subnetworks under attack and (n) disabling vulnerable network protocols — that can be used to contain T1008 by isolating systems using fallback channels or disabling the protocols carrying them; however, the control is primarily a preventive/detective network-security baseline, and the actual incident-response process (assessment, escalation, closure) belongs to A.5.26, so the response contribution is a slice of the control rather than its core function.
- T1016partialdetects — A.8.20's network-focused logging and monitoring (clause f) and connection detection (clause j) can surface the network-traffic-bearing portions of T1016 — ARP sweeps, remote Network Device CLI queries, and routing discovery that traverse the network — but the bulk of the technique is local command execution (ipconfig/ifconfig/route/nbtstat on a compromised host), which is host telemetry's job and sits outside A.8.20's network vantage point.
- T1016.001partialdetects — A.8.20(f) requires logging and monitoring to enable recording and detection of actions relevant to information security, which can surface network connectivity checks, but the control's scope is set by implementation rather than mandating the depth of instrumentation that would catch every method.
- T1018partialprevents — Network segmentation, filtering and isolation (items i, j, m) prevent cross-segment discovery vectors like ping and net view from reaching other subnets, but local/passive vectors — ARP cache, hosts file, CDP neighbors on a network device the adversary already controls — remain available regardless of network controls, so a substantial slice of the technique's mechanism is unaffected.
- T1018partialresponds — A.8.20 has response capabilities against T1018 via network isolation (m), firewall filtering (i), and network access control (j), which can contain the discovery activity by isolating the source or blocking the probes. However, A.8.20 does not perform eradication, investigation, or evidence collection — those belong to A.5.26/A.5.28. The response reaches the contact (the probes hitting the estate) but not the full scope of a complete response, similar to the A.5.26 vs T1595 anchor.
- T1020partialdetects — A.8.20 item f) explicitly requires logging and monitoring to detect actions affecting or relevant to information security, which surfaces automated exfiltration in flight on the organization's networks (especially via boundary or anomalous traffic patterns); this is bounded by the control's network-scope focus, leaving non-network or pre-transmission collection undetected.
- T1020partialprevents — A.8.20's firewalls, connection filtering, protocol disabling and network segregation can block some outbound exfiltration channels, but T1020's automation can ride any reachable path (C2, alternative protocols, DNS, etc.) and the clause does not address the automation mechanism itself, leaving a broad remainder of channels the control does not reach.
- T1020partialresponds — A.8.20 includes explicit response mechanisms — drawbridges for temporarily isolating subnetworks under attack (m), firewall rules to block exfiltration channels (i), and filtering/restricting system connections (j) — that can contain automated exfiltration once detected, but the control is primarily preventive/detective and the full incident response procedure (assessment, evidence collection, eradication, closure) lives in A.5.26, so what A.8.20 contributes is network-level containment rather than complete response.
- T1020.001partialprevents — A.8.20's hardening (k), authentication (h), and connection filtering (i) directly prevent the malicious-modification vector (e.g., ROMMONkit, Patch System Image enabling rogue mirroring), but the technique also abuses legitimately configured traffic mirroring — a native feature on network devices and cloud platforms — where the mirror destination itself is the adversary's target or the management plane is compromised via legitimate credentials; A.8.20 does not reach those paths, and cloud-side mirroring configuration (AWS/GCP/Azure) sits largely outside its scope.
- T1021partialprevents — Network segmentation, connection filtering, and device authentication limit an adversary's ability to discover and directly interact with remote services for lateral movement.
- T1021partialdetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16) plus detection-oriented items (j: detecting device connections; m: isolating subnetworks under attack; n: disabling vulnerable protocols), which surface T1021's use of remote protocols (SSH/RDP/VNC) on the network.
- T1021.001partialdetects — A.8.20's network monitoring (clause f) and connection detection (clause j) can see RDP sessions and flag anomalies in source, destination, timing or volume, but the technique's defining feature is use of valid accounts, so a session from a normal source with legitimate-looking traffic is indistinguishable from authorized use at the network layer alone — credential legitimacy and post-authentication activity are identity/host concerns outside A.8.20's scope.
- T1021.001partialprevents — A.8.20's network controls (firewalls restricting/filtering connections, network segmentation, disabling vulnerable protocols, isolating critical subnetworks) can prevent RDP from being reachable in many configurations, but where RDP is legitimately enabled for remote administration and valid credentials exist, the network control alone does not stop the technique — the success then depends on credential validity, which is outside A.8.20's scope.
- T1021.001partialresponds — A.8.20 contains a real response capability through (m) temporarily isolating critical subnetworks under attack and (i) filtering/restricting connections, which can terminate an active RDP session at the network boundary; however, the clause is primarily preventive and its response reach is limited to network-layer containment — host-level eradication, evidence collection, and incident assessment are outside its scope, so it performs a slice of the response tasks rather than the bulk.
- T1021.002partialdetects — A.8.20 item (f) requires logging and monitoring of network activity (cross-referencing A.8.16/A.8.15), and item (j) explicitly mandates detecting connections to the network — SMB admin share usage generates network traffic that falls within that scope. Partial because the clause's primary purpose is protection rather than detection, the monitoring scope is set by implementation rather than mandated at a depth that specifically surfaces lateral SMB movement, and detection of valid-account SMB use depends on correlating authentication context with share access, which the clause does not require.
- T1021.002partialprevents — A.8.20's network controls (firewall filtering, segmentation, disabling vulnerable protocols, segregating admin channels) can block lateral SMB movement between network segments and restrict where admin shares are reachable, but SMB is typically required for legitimate file sharing and administration within a domain, and the technique's core mechanism — valid accounts authenticating to admin shares — is outside A.8.20's scope, so what is prevented is a slice of the attack surface rather than the technique.
- T1021.002partialresponds — A.8.20 contains response capabilities against lateral SMB movement via item (m) drawbridge isolation of subnetworks under attack and item (i) firewall reconfiguration to block SMB traffic, which are real containment actions; however, the clause does not address the compromised source host, the adversary's authenticated session, or credential remediation, so containment of spread is partial while eradication is largely outside its scope.
- T1021.003partialdetects — A.8.20 item (f) requires network-level logging and monitoring to detect actions affecting information security, and DCOM lateral movement generates observable RPC traffic (port 135, dynamic ports) that network monitoring can see; however, the clause is about network security broadly and does not specifically target DCOM, so distinguishing malicious DCOM lateral movement from legitimate COM/RPC usage requires deeper analysis the clause does not mandate.
- T1021.003partialprevents — A.8.20's network controls (firewalls filtering system connections, network segmentation, disabling vulnerable protocols) can block DCOM/RPC traffic across network boundaries, but DCOM is a built-in Windows protocol often required for legitimate operations within a network segment, and the technique also depends on valid accounts which is outside A.8.20's scope.
- T1021.003partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which provides network-level containment of DCOM lateral movement, but earns partial because it does not eradicate the adversary's foothold or stop intra-subnet execution.
- T1021.004partialdetects — A.8.20 items (f) and (j) place SSH connections and authentication events within the network monitoring and connection-detection scope, so the session establishment is observable; however, the technique uses valid accounts and an encrypted tunnel, so the malicious intent is not directly visible at the network layer and behavioural or post-authentication telemetry is needed to distinguish adversarial SSH from legitimate use.
- T1021.004partialprevents — A.8.20's firewalling, network segmentation, disabling of vulnerable protocols, and isolation of critical subnetworks can prevent SSH from being reachable in many configurations, but SSH is often a required service and where it is legitimately enabled the network-level controls restrict rather than prevent the technique, especially since T1021.004 uses valid accounts that pass network authentication.
- T1021.004partialresponds — A.8.20(m) explicitly addresses responding to a network attack in progress by temporarily isolating critical subnetworks, which is a genuine containment action against SSH-based lateral movement; however, the clause cannot kill the active session, eradicate the foothold on the already-authenticated host, or remove the adversary's credentials, so the response is a network-level slice rather than the full containment-and-eradication set.
- T1021.005partialprevents — A.8.20's firewall restrictions (i), network segmentation (l, m), authentication of network connections (h), and disabling of vulnerable protocols (n) can block or constrain VNC traffic and limit lateral movement via RFB, but VNC is a legitimate remote administration tool that organizations commonly permit, and the technique explicitly uses valid accounts — so where VNC is allowed for legitimate use, A.8.20 does not prevent its abuse.
- T1021.006partialdetects — A.8.20 requires logging and monitoring of network actions, which can surface anomalous WinRM traffic patterns, but the control's scope is network-level and does not mandate host-level instrumentation to detect the in-process activity that follows a successful WinRM session.
- T1021.006partialprevents — A.8.20's firewalls, network segmentation, protocol disabling, and connection filtering can block WinRM ports (5985/5986) between segments and restrict where remote management traffic is permitted, but the technique uses valid accounts and WinRM is often required for legitimate administration, so within allowed network paths the technique still runs.
- T1021.006partialresponds — A.8.20 includes response-capable elements — item (m) temporarily isolating critical subnetworks during an attack and item (n) disabling vulnerable network protocols — that can contain and eradicate an active WinRM-based intrusion, but these are a subset of a control whose primary purpose is preventive, so the response share is a slice rather than the bulk.
- T1021.007partialprevents — A.8.20's network controls (segmentation, egress filtering, restricting system connections) can constrain which compromised endpoints reach cloud service APIs, but the technique's core mechanism is valid-credential authentication to cloud services over standard HTTPS — an identity-layer issue that network-layer controls do not address, so what is prevented is a slice of reachable paths rather than the technique itself.
- T1021.008partialdetects — A.8.20's network monitoring (item f) and connection detection (item j) can observe the outbound API calls to cloud provider endpoints (Azure Serial Console, EC2 Instance Connect, SSM), but the actual VM console access and privileged actions execute within the cloud provider's infrastructure outside the organization's monitored network, and the connections use valid authentication so they resemble legitimate cloud management traffic.
- T1021.008partialprevents — A.8.20's network controls (firewalls, filtering, restricting connections per items i/j, hardening per k, disabling vulnerable protocols per n) can constrain some network-level paths to cloud VMs, but T1021.008 specifically leverages cloud provider control-plane APIs (Azure Serial Console, EC2 Instance Connect, SSM) that authenticate via cloud IAM and often bypass traditional network-layer controls, so the technique's primary vector — valid accounts through cloud-native methods — sits outside what network security can reach.
- T1021.008partialresponds — Temporarily isolating critical subnetworks contains the adversary's ability to pivot from the compromised VM, though it does not eradicate their underlying cloud control plane access.
- T1027partialdetects — A.8.20's network monitoring (item f) and traffic filtering can surface some in-transit obfuscation and anomalous protocol patterns, but T1027 is explicitly designed to evade detection and the bulk of the technique operates at the file/system level on endpoints, which is outside A.8.20's network vantage point.
- T1027.006partialdetects — A.8.20's network monitoring (item f) can observe the HTML file traversing the network and potentially flag some indicators (unusual sizes, known patterns), but HTML smuggling is specifically engineered to evade network-level content filters — the file presents as benign text/html and the malicious payload is constructed client-side in the browser via JavaScript Blobs or Data URLs, where the network control has no visibility.
- T1027.017partialdetects — A.8.20's network logging and monitoring (item f) can observe downstream network effects of SVG smuggling — the download of assembled payloads, C2 connections, or redirects to malicious sites — but the core technique is specifically designed to evade content inspection by hiding payloads inside files that appear to be benign images; network-level firewalls and traffic monitoring do not deeply inspect SVG/XML content for embedded scripts, so the smuggling mechanism itself passes through largely unseen.
- T1029partialdetects — A.8.20's network monitoring (item f) can surface some scheduled transfers through traffic analysis, but T1029 is specifically designed to blend with normal activity patterns, and the clause's monitoring is general-purpose rather than tailored to timing-based exfiltration, leaving a substantial remainder undetected.
- T1030partialdetects — A.8.20(f) explicitly mandates logging and monitoring to enable recording and detection of actions relevant to information security, and T1030 operates entirely at the network layer within that scope, so the traffic is observable; however, the technique is specifically designed to evade threshold-based alerts by staying below them, and A.8.20 sets scope rather than mandating the behavioural or volumetric analysis needed to reliably surface chunked exfiltration, leaving a nameable remainder.
- T1030partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which provides a network-level containment mechanism to halt ongoing exfiltration, though it does not eradicate the adversary's foothold.
- T1033partialdetects — A.8.20's logging and monitoring scope (item f) and network device hardening (item k) cover the network device CLI portion of T1033 (e.g., `show users`, `show ssh` on routers/switches), but the bulk of the technique runs as host-level commands (`whoami`, `w`, `who`, environment variables) on workstations and servers, which sit outside A.8.20's network-focused monitoring scope.
- T1037partialdetects — A.8.20's monitoring scope (item f) covers network traffic and network-relevant actions, which can surface some manifestations of T1037 — particularly scripts that beacon out, contact internal logging servers, or generate anomalous network behavior at boot/logon, and the subset running on network devices. The core technique (script execution at boot/logon on endpoints) is an endpoint event outside A.8.20's primary vantage point, so what is detected is a slice rather than the technique itself.
- T1037.003partialdetects — A.8.20(f)'s network logging and monitoring can surface some indirect indicators of T1037.003 — SMB access to script shares, traffic generated by script execution, AD replication traffic on GPO changes — but the core mechanism (AD/GPO assignment of the logon script and its endpoint execution at logon) is an AD/host event outside A.8.20's network-focused scope, so detection is a slice rather than coverage of the technique.
- T1039partialdetects — A.8.20's logging and monitoring (item f) can record network share access traffic and surface anomalous patterns (volume, source, timing), but the technique uses valid credentials on a compromised system and looks like legitimate file access, so network-level monitoring alone cannot reliably distinguish malicious from authorized share reads without behavioral baselines or content awareness.
- T1039partialprevents — A.8.20's network segmentation (m, l), connection filtering (i, j), and system authentication (h) can prevent a compromised system from reaching specific network shares, but the technique uses legitimate file-sharing protocols that must remain functional for business operations, so shares the compromised host is authorized to access remain reachable and the technique still runs against them.
- T1039partialresponds — The control includes temporarily isolating critical subnetworks during an attack, which contains the adversary's ability to access network shared drives across subnets, though it does not eradicate their presence on the compromised host.
- T1040partialdetects — A.8.20's detection mechanisms — (f) logging/monitoring of actions affecting security, (j) detecting connection of equipment and devices, and (c) configuration-file documentation — can surface some indicators (unauthorized device attachment, span-port or traffic-mirroring configuration changes, cloud mirroring API calls), but the core of T1040 is passive capture on an already-connected, already-compromised host, which generates no network-visible signal and remains largely invisible to network-layer controls.
- T1040partialprevents — A.8.20 prevents some instances of T1040 via disabling vulnerable protocols (n), hardening network devices against span port misuse (k), authenticating and restricting equipment connection (i, j), and segmenting networks (l, m), but it does not mandate encryption of traffic and cannot prevent an adversary with a foothold on a segment from running the technique against whatever cleartext or metadata remains — the technique itself is not stopped, only its surface is reduced.
- T1040partialresponds — A.8.20 includes genuine response mechanisms — (m) drawbridges to isolate subnetworks under attack and (j) restricting/authenticating devices to remove rogue sniffing equipment — but the clause is primarily a preventive and detective network-security control, not an incident response procedure; the response elements are a slice of its scope rather than its purpose.
- T1041partialdetects — A.8.20 item f explicitly requires logging and monitoring to detect actions affecting or relevant to information security, which surfaces anomalous exfiltration over an established C2 channel (especially when it deviates from baseline traffic patterns); however, the clause's scope is limited to the organization's own networks and does not guarantee detection of all protocol-conforming exfiltration (e.g. encrypted C2 blending with legitimate traffic or activity outside monitored segments).
- T1041partialprevents — A.8.20's network controls (firewalls, connection filtering, system authentication, disabling vulnerable protocols) can prevent some C2 channels from being established in the first place, but T1041 specifically uses an already-established C2 channel with an allowed protocol, so exfiltration over that channel is largely outside the control's prevention reach.
- T1041partialresponds — A.8.20 includes explicit response-capable items — (m) temporarily isolating critical subnetworks under attack and (i) restricting/filtering connections — that can contain T1041 by severing the C2 channel or quarantining the affected segment once exfiltration is underway, but it is a technical slice of response (containment via network means) and does not perform eradication of the foothold, evidence collection, assessment, or closure, which belong to A.5.26.
- T1046partialprevents — A.8.20's firewalls, filtering, isolation, and protocol disabling prevent successful enumeration of many internal services, but the scan still runs against public-facing services that must remain accessible, against any service the firewall permits through, and within the internal network where lateral scanning is not blocked — so the technique is constrained, not stopped.
- T1046partialresponds — A.8.20 includes explicitly responsive elements — (m) temporarily isolating critical subnetworks under attack via drawbridges, (j) restricting connections of equipment/devices, and (i) firewall filtering that can block an active scanning source — so it can contain an ongoing T1046 scan; however, it is primarily a preventive/architectural control, performs no assessment or evidence collection, and has limited eradication reach against an external scanner, leaving the response a slice of the full verb.
- T1047partialdetects — A.8.20's network-level monitoring (clause f) and connection detection (clause j) can surface remote WMI traffic on ports 135 (DCOM), 5985/5986 (WinRM), but local WMI execution produces no network traffic and is invisible to network controls, so only the remote variant is caught.
- T1047partialprevents — A.8.20's firewalling, port filtering, and network segregation can block remote WMI abuse (DCOM port 135, WinRM ports 5985/5986), but WMI is a built-in Windows administration feature that runs locally on every Windows host, and local WMI execution — including the canonical `wmic.exe shadowcopy delete` — is unaffected by network controls.
- T1047partialresponds — A.8.20(m) explicitly provides for temporarily isolating critical subnetworks (drawbridges) when the network is under attack, which is a genuine containment action against an in-progress WMI-based attack — particularly the remote-WMI lateral-movement variant over DCOM/WinRM. However, this is one slice of the response set: A.8.20 does not kill the WMI process, remove the payload, collect forensic evidence, or restore affected hosts, and WMI execution is largely host-based, so network-level isolation reaches only the lateral-spread component.
- T1048partialdetects — A.8.20 requires logging and monitoring to enable recording and detection of actions affecting information security, which covers network traffic including exfiltration over alternative protocols, but the clause sets a requirement rather than mandating specific detection depth, so an implementation monitoring only standard protocols is fully conformant and misses exfiltration over non-standard channels.
- T1048partialprevents — A.8.20's firewalls (i), disabled vulnerable protocols (n), and subnetwork isolation (m) can block some exfiltration channels, but the commonly used protocols (HTTPS, DNS, SMTP) are business-essential and must remain permitted, so the control prevents a slice of the technique's channels rather than the bulk.
- T1048partialresponds — A.8.20 provides network-level containment capabilities — item (m) isolating subnetworks during an attack, item (n) disabling vulnerable protocols, item (i) filtering connections — that can act against an exfiltration already underway, but it does not perform the full set of response tasks (no incident assessment, no evidence collection, no eradication of the actor's foothold), so it is a slice of the response set rather than the bulk.
- T1048.001partialprevents — A.8.20's network-level mechanisms (firewalls, filtering, segmentation, protocol restrictions) can prevent some exfiltration by blocking unauthorized destinations and channels, but symmetric encryption defeats content inspection and the technique can ride any permitted protocol, leaving a substantial remainder.
- T1048.002partialdetects — A.8.20 item (f) explicitly mandates logging and monitoring of network actions relevant to information security, and items (i)/(j) provide filtering and connection-detection mechanisms that can surface anomalous outbound flows; however, T1048.002 uses standard asymmetrically-encrypted protocols (TLS/HTTPS) whose payloads are opaque to network inspection, so detection reduces to traffic-volume, destination-reputation and timing analysis — a slice of the technique rather than a bounded remainder of a working mechanism.
- T1048.002partialprevents — A.8.20's network controls (firewalls, segmentation, filtering of destinations, isolation of critical subnetworks) can restrict the paths and endpoints available for exfiltration, but T1048.002 uses legitimate asymmetrically encrypted protocols (HTTPS/TLS) that must be permitted for business operations, so the network cannot inspect or block the content; the preventive effect is real but bounded to path/destination restriction, leaving the technique reachable over any allowed encrypted channel.
- T1048.002partialresponds — A.8.20 has explicit response mechanisms — (m) temporarily isolating critical subnetworks via drawbridges if the network is under attack, and (i) restricting/filtering system connections — that can act on an ongoing exfiltration. However, A.8.20 is primarily a preventive network-security control rather than a dedicated incident-response procedure, and T1048.002's use of asymmetrically encrypted protocols (HTTPS/TLS) makes the exfiltration traffic indistinguishable from legitimate encrypted flows, limiting what filtering and isolation can confirm or fully eradicate without decrypting content.
- T1048.003partialprevents — A.8.20's firewall/filtering (i), protocol disabling (n), and connection restriction (j) can block some exfiltration channels, but T1048.003 deliberately uses protocols (HTTP, DNS, FTP) that are typically permitted for legitimate business traffic, so the control cannot prevent the technique's full denominator.
- T1048.003partialresponds — A.8.20 provides network-level containment mechanisms that can act once exfiltration is underway — item (m) isolating subnetworks under attack, item (i) firewall filtering to block egress, item (n) disabling vulnerable protocols — but these are pre-configured network controls, not an incident response procedure; they contain the channel without eradicating the actor, collecting evidence, or performing the other response tasks, so the verb holds only partially.
- T1049partialdetects — A.8.20 item (f) requires logging and monitoring of actions affecting information security, and item (k) covers hardening (including logging) of network devices, so queries against network infrastructure are detectable; however, the bulk of T1049 is host-level process execution (netstat, lsof, net session) which sits outside A.8.20's network-focused scope and is primarily the province of host telemetry controls.
- T1053.007partialdetects — A.8.20 explicitly covers virtualized networks and SDN, so container/Kubernetes network traffic is in scope; items (f) logging/monitoring and (j) detecting connections can surface anomalous API calls to the Kubernetes control plane and unusual inter-pod or egress traffic from scheduled containers, but the core mechanism — CronJob scheduling and in-container code execution — is not a network event and remains invisible to network-layer monitoring alone.
- T1056partialdetects — A.8.20's network logging and monitoring (item f) can surface some indicators of input capture — outbound exfiltration of captured credentials, network-based credential capture, or anomalous connections from a host running a keylogger — but the core mechanism of T1056 (API hooking, keystroke logging, fake portal dialogs) is endpoint-resident and largely outside a clause whose monitoring scope is explicitly the network.
- T1056.003partialdetects — A.8.20's item (f) explicitly invokes network-level monitoring and logging (cross-referencing A.8.16/A.8.15), which can surface the credential-exfiltration traffic from a compromised portal to an external collector; however, the core mechanism — code installed on the web portal capturing credentials at the application layer before login — is largely invisible to network monitoring, so what is caught is a slice (the egress) rather than the technique itself.
- T1056.003partialprevents — A.8.20's network-level controls (firewalls, network segmentation, hardening, disabling vulnerable protocols) can help prevent initial compromise of the externally facing web server, but the technique explicitly operates post-compromise via legitimate administrative access or application-layer exploitation, where network security controls do not stop modification of the login page itself.
- T1059partialdetects — A.8.20's logging and monitoring scope (item f) and connection detection (item j) can surface network artifacts of T1059 — downloaded payloads, C2 callbacks, remote-execution traffic — but the clause's scope is the network, not host process execution, so local interpreter activity without network involvement is not observed.
- T1059.001partialdetects — A.8.20's logging/monitoring scope (item f) and connection filtering (items i, j) can surface the network-visible portion of PowerShell activity — remote sessions, payload downloads, C2 callbacks — but local script execution, in-memory invocation, and .NET assembly loading produce no network traffic for these controls to see.
- T1059.003partialdetects — A.8.20's monitoring scope (item f) and connection-detection (item j) can surface network artifacts of cmd.exe abuse — remote shell sessions, C2 channels, traffic from cmd-spawned processes — but the core technique is local execution on a Windows host, which a network-focused control does not see.
- T1059.008partialprevents — A.8.20's mechanisms (h authentication, i/j filtering, k hardening, l segregating admin channels, n disabling vulnerable protocols) directly constrain the attack surface for network device CLI abuse, but the CLI is a required function for legitimate administration and cannot be removed, so an adversary with valid access can still execute the technique.
- T1059.009partialdetects — A.8.20's logging and monitoring of network actions (clause f) and connection detection (clause j) can surface metadata about cloud API calls — source systems, destinations, volumes, timing — enabling behavioural anomaly detection against T1059.009, but the malicious commands themselves travel inside encrypted HTTPS payloads to legitimate cloud endpoints and are invisible to network-layer controls, so what is detected is a slice of behavioural signals rather than the technique's substance.
- T1059.012partialdetects — A.8.20's network logging and monitoring (clause f) and admin-channel segregation (l) can surface some hypervisor management traffic (SSH to ESXi, vCenter connections), but the technique's core is local command execution on the hypervisor host via esxcli/vim-cmd, which is largely outside a network-security control's vantage point.
- T1059.013partialdetects — A.8.20's network logging and monitoring (item f) and connection authentication (items h, j) can surface remote Docker daemon or Kubernetes API server calls traversing the network, but a substantial share of T1059.013 — local docker CLI invocations, docker socket access, and kubectl with a local kubeconfig — produces no network traffic for A.8.20 to observe.
- T1059.013partialprevents — A.8.20's network controls (firewalls, network segmentation, authentication of systems on the network, restricting/filtering connections) prevent the network-based vector of T1059.013 — unauthorized access to exposed Docker daemons or Kubernetes API servers — but cannot prevent execution via local sockets, from hosts already authenticated on the management network, or from within containers that already have CLI access, leaving a meaningful slice of the technique reachable.
- T1069partialdetects — A.8.20's network logging and monitoring (item f) can record LDAP/Kerberos-based group enumeration against domain controllers, but the bulk of T1069 — local `net group`/`net localgroup`, PowerShell, WMI, and BloodHound collection against local SAM/LSASS — happens on-host and is outside the network-security control's vantage point.
- T1069.002partialdetects — A.8.20 clause (f) requires logging and monitoring of network actions relevant to information security, and T1069.002's enumeration queries (LDAP, SMB/RPC to domain controllers) traverse the network and are observable at that vantage point; however, the clause sets scope rather than mandating detection of this specific technique, the queries use standard protocols that resemble legitimate admin traffic, and detection depends on implementation specifics such as anomaly detection or targeted signatures.
- T1070partialdetects — A.8.20's monitoring scope (item f) and network-device hardening (item k) cover the network-device portion of T1070 — clearing router/switch logs or modifying network device artifacts — but the bulk of T1070 is host-level (command histories, file metadata, system logs on endpoints/servers), which sits outside A.8.20's network-focused mechanism.
- T1070.003partialdetects — A.8.20(f) covers logging and monitoring of network devices, and (k)/(l) cover hardening and segregating network administration channels, so clearing command history on a network device via `clear history` is within the clause's monitoring scope; however, the bulk of T1070.003 — Linux/macOS bash_history, Windows PSReadLine history, and ESXi shell.log clearing — is host-local file activity that network security controls have no vantage point to observe.
- T1071partialdetects — A.8.20(f) explicitly mandates logging and monitoring to enable detection of actions affecting information security, and (j) explicitly includes detecting connections of equipment to the network, so the control provides a detection framework for network-layer activity; however, T1071 is specifically designed to blend in with legitimate application layer traffic (HTTP, DNS, SMB, etc.), and the control's primary focus is on protection mechanisms (filtering, segregation, hardening) rather than deep protocol-content inspection, so what is surfaced is a slice of the technique's activity rather than the bulk.
- T1071partialprevents — A.8.20's filtering, protocol disabling, network segmentation and egress controls can block some application-layer C2 channels, but T1071 is specifically designed to use protocols (HTTP/S, DNS, SMTP, etc.) that must remain allowed for business function, so the technique still runs over permitted protocols to permitted destinations.
- T1071partialresponds — A.8.20 item (m) explicitly provides for temporarily isolating critical subnetworks when the network is under attack, and item (n) disabling vulnerable protocols can be deployed as eradication of an active C2 channel — both are genuine response actions against T1071 once underway. But the clause is primarily a preventive/detective network-security control, and its response surface is limited to network-level containment rather than the full incident-response set (assessment, evidence, eradication of the foothold), so the verb holds only partially.
- T1071.001partialprevents — A.8.20's filtering, firewalling, network segmentation and protocol restrictions can constrain where and how web-protocol C2 traffic flows, but HTTP/S are essential to operations and cannot be blocked wholesale, and the technique is explicitly designed to blend with allowed traffic — so what is prevented is a slice (egress to known-bad destinations, lateral C2 across segmented boundaries) rather than the bulk.
- T1071.001partialresponds — A.8.20 includes explicit response capabilities — drawbridges to isolate subnetworks under attack (m), firewall reconfiguration to block C2 traffic (i), and disabling vulnerable protocols (n) — but these are a subset of the clause whose bulk is preventive, and detection of the web-protocol C2 channel belongs to A.8.16 monitoring rather than this control.
- T1071.002partialprevents — A.8.20's firewalling (i), protocol disabling (n), system authentication (h), and network segmentation (m) can block or restrict file-transfer protocols at boundaries and between segments, but T1071.002 deliberately uses legitimate, commonly-required protocols (SMB, FTP) to blend with normal traffic, so the control cannot eliminate the technique without breaking business functionality.
- T1071.002partialresponds — A.8.20 item (m) — temporarily isolating critical subnetworks via drawbridges if the network is under attack — is a genuine containment action against an adversary abusing file transfer protocols for C2, and item (n) disabling vulnerable protocols can be applied mid-incident; but the clause's primary function is preventive (firewalls, filtering, hardening, authentication per items h/i/j/k), and the response slice is narrow — it isolates subnetworks, not the actor's foothold, and does not eradicate, collect evidence or close the incident.
- T1071.003partialprevents — A.8.20's network controls (firewall filtering, protocol restriction, network segmentation, disabling vulnerable protocols) can constrain mail-protocol C2 by restricting which systems may initiate SMTP/POP3/IMAP and to which destinations, but mail protocols are business-essential and the technique is designed to blend with legitimate traffic, so the control cannot eliminate the channel.
- T1071.003partialresponds — A.8.20 item (m) explicitly enables temporarily isolating critical subnetworks (e.g. with drawbridges) when the network is under attack, which is a genuine containment action against mail-protocol C2; however, the clause is primarily preventive (filtering, hardening, protocol disabling) and detective (logging, monitoring), and it does not perform assessment, evidence collection, eradication, communication or closure — so the response it delivers is a slice of the full response procedure.
- T1071.004partialresponds — A.8.20's clause (m) — temporarily isolating critical subnetworks with drawbridges if the network is under attack — is an explicit response action that can contain DNS tunneling by severing the C2 channel, but the clause is one item among fourteen and addresses isolation rather than the full containment-plus-eradication that `responds` asserts; the rest of A.8.20 is preventive (firewalls, filtering, hardening, protocol disabling) or detective (logging, monitoring), so the response share is a slice.
- T1071.005partialdetects — A.8.20's logging/monitoring (f) and connection-detection (j) clauses provide a framework that can surface unauthorized pub/sub brokers, anomalous pub/sub volumes, or known-bad signatures, but T1071.005 is specifically designed to blend in with legitimate MQTT/XMPP/AMQP/STOMP traffic, and the clause sets scope rather than mandating the deep protocol inspection or behavioral baselining needed to catch abuse of authorized brokers.
- T1071.005partialprevents — A.8.20's firewall filtering (i) and disabling of vulnerable protocols (n) can block pub/sub traffic where the organization does not legitimately use it, but the technique is specifically designed to blend in with existing pub/sub traffic, so where MQTT/XMPP/AMQP/STOMP are in legitimate use the control cannot prevent the C2 channel from operating.
- T1071.005partialresponds — A.8.20 item (m) explicitly provides for temporarily isolating critical subnetworks (drawbridges) when the network is under attack, which is a genuine containment action against ongoing C2 over pub/sub protocols; however, this is one sub-item in a clause primarily oriented to prevention (filtering, authentication, hardening, disabling protocols), and the clause does not address eradication of the foothold or the compromised broker.
- T1072partialdetects — A.8.20 item (f) explicitly requires logging and monitoring of network actions, and items (i)-(j) provide for filtering and detecting connections, so network-level telemetry can surface some indicators of T1072 abuse (anomalous traffic from management servers, C2 callbacks via web protocols, unusual lateral patterns); however, the technique's core is abuse of legitimate enterprise tools with valid credentials, which produces traffic that looks like normal administrative activity and is largely invisible to network-layer monitoring alone.
- T1072partialprevents — A.8.20's network controls — segmentation of administration channels (l), firewall filtering of system connections (i), isolation of critical subnetworks (m), and authentication of systems on the network (h) — can constrain the reach of a compromised deployment tool and limit lateral movement, but the technique still functions on systems the tool legitimately reaches, and A.8.20 does not address the initial compromise of the deployment system itself.
- T1072partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which provides network-level containment that bounds the adversary's ability to push commands or move laterally via centralized deployment tools, though it does not eradicate their foothold on the deployment server itself.
- T1074partialdetects — A.8.20's logging/monitoring (item f) and traffic filtering can surface staging that traverses network segments or uses network shares, but the core of T1074 — copying files into a local directory or within a cloud instance — is a host/file-system activity largely outside the clause's network vantage point.
- T1074.002partialdetects — A.8.20(f) requires logging and monitoring of network actions, and remote data staging across systems generates network traffic (SMB, RDP, file-copy protocols) that flow monitoring and connection logging can surface; however, the clause's primary purpose is network access control and protection rather than detection, local staging within a single host produces no network signal, and small or protocol-conforming transfers evade the monitoring the clause mandates, so detection is a slice rather than a bounded remainder of a working mechanism.
- T1074.002partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which contains the network-based movement of data to the remote staging location, though it does not eradicate the data already staged.
- T1078partialdetects — A.8.20(f) logging/monitoring of network actions can surface behavioural anomalies from valid account abuse (unusual connection patterns, lateral movement, access from atypical sources), but the core network security mechanisms — firewalls, filtering, segmentation (i, j, l, m) — see the traffic as legitimate authenticated sessions and do not detect the abuse; detection is limited to the slice that network-level monitoring can observe, not the credential misuse itself.
- T1078partialprevents — A.8.20's network-level controls (segmentation per item m, connection restriction/filtering per items i and j, system authentication per item h) can constrain where and how valid credentials are exercised and limit lateral movement, but the technique's defining feature is that the credentials are legitimate and pass network authentication, so the clause does not prevent the abuse itself — that is A.8.5's surface.
- T1078partialresponds — The control provides the capability to temporarily isolate critical subnetworks when under attack, which partially contains an adversary's ability to pivot or access resources using compromised valid accounts, though it does not eradicate the compromised credentials themselves.
- T1078.001partialprevents — A.8.20's network controls (firewalls, filtering, network authentication, segmentation) can block remote access paths that would abuse default accounts, but they do not remove the accounts themselves and cannot prevent local abuse or access via paths the network permits — the underlying account-existence problem belongs to A.8.9.
- T1078.002partialdetects — A.8.20's network-level logging/monitoring (f), device-connection detection (j), and traffic filtering (i) can surface some indicators of domain account abuse — unusual source IPs, new devices authenticating, lateral-movement traffic patterns — but the core abuse of valid domain credentials is an authentication/identity-layer event that looks like legitimate traffic at the network boundary, so what is caught is a slice of network-visible indicators rather than the technique itself.
- T1078.003partialdetects — A.8.20's logging/monitoring (item f), system authentication (item h), and connection detection (item j) can surface network-visible aspects of local account abuse — lateral movement using reused local credentials across systems, network logons with local accounts, and anomalous authentication patterns — but the core abuse (direct console logon, local privilege escalation, credential harvesting on a single host) is a host-level event outside the network control's vantage point.
- T1078.003partialprevents — A.8.20's network segmentation (item m), filtering (item i), and system authentication (items h, j) can limit the lateral-movement component of T1078.003 — the technique's own description names password reuse across machines on a network — but the control does not prevent the local account from existing or being abused on the system where its credentials reside, which is the core of the technique.
- T1080partialdetects — A.8.20's monitoring scope (item f) covers logging and monitoring of actions relevant to information security on the network, which can surface some indicators of T1080 such as unusual SMB access patterns and lateral movement to shared drives, but the core tainting action — modifying or appending malicious content to files on shared storage — is a file-system event that network-level monitoring does not directly observe.
- T1080partialprevents — A.8.20's network segmentation (m), system authentication (h), and connection filtering (i,j) can limit which systems reach shared storage locations and constrain lateral movement to shares, but the control operates at the network layer and does not prevent an already-authorized system from writing malicious content to a network drive or repository it has access to — the core tainting mechanism is a file/share-level operation outside A.8.20's scope.
- T1082partialdetects — A.8.20's network-level logging and monitoring (item f) and network device access controls (items h, j, k) can detect the network-mediated variants of T1082 — network device CLI queries (show version) and cloud provider API calls — but the bulk of the technique is host-local execution (systeminfo, systemsetup, esxcli) which falls outside A.8.20's network scope.
- T1087partialdetects — A.8.20's network monitoring (item f) and connection controls can surface account discovery that traverses the network — LDAP/SMB enumeration against domain controllers, cloud identity-provider API calls — but local enumeration via PowerShell, `net` commands, or reading local files produces no network traffic the clause's mechanism observes, so a meaningful slice of the technique is invisible.
- T1087.002partialdetects — A.8.20 item (f) mandates network-level logging and monitoring for actions relevant to information security, and domain account enumeration (LDAP, SMB/RPC, Kerberos traffic to domain controllers) traverses the network where such monitoring can observe it; however, the control's primary purpose is network protection (firewalls, segmentation, hardening per items i, k, l, m, n), and authenticated enumeration queries often resemble legitimate admin traffic, so detection depends on implementation depth rather than being inherent to the clause.
- T1087.004partialdetects — A.8.20's network monitoring (item f) can observe metadata of API calls traversing the network to cloud providers — source, destination, volume, timing — enabling some anomaly-based detection of enumeration activity, but the API call content is encrypted HTTPS and indistinguishable from normal cloud service usage, so the meaningful detection (which specific enumeration commands ran) belongs to cloud audit logs outside this control's scope.
- T1090partialdetects — A.8.20 item (f) explicitly requires logging and monitoring of network actions, which can surface proxy traffic patterns, anomalous routing, and C2 communications, but sophisticated proxy use through legitimate services (CDNs, cloud providers) and encrypted proxy chains are designed to look like normal traffic and the clause sets monitoring requirements rather than mandating specific detection depth for proxy traffic.
- T1090partialprevents — A.8.20's sub-controls (egress filtering, disabling vulnerable protocols, network segmentation, restricting connections) can prevent some proxy implementations and constrain where proxies operate, but proxies fundamentally use legitimate network functionality (HTTPS, allowed services) that must remain available for business operations, so the technique cannot be prevented entirely.
- T1090partialresponds — A.8.20 includes response-capable elements — temporarily isolating critical subnetworks under attack (m), restricting and filtering system connections (i), and disabling vulnerable network protocols (n) — which can contain T1090 proxy traffic once detected, but the clause is primarily preventive and does not address eradication of the proxy infrastructure, assessment, or evidence collection.
- T1090.001partialprevents — A.8.20's network segmentation (items l, m), filtering (item i), and authentication of systems on the network (item h) can prevent some internal proxying by restricting which compromised systems can communicate, but the technique explicitly rides over existing trusted communications paths and common protocols like SMB, so where systems legitimately need to communicate the proxy channel survives — a slice prevented, not the bulk.
- T1090.001partialresponds — A.8.20 includes direct response-capable elements — (m) temporarily isolating critical subnetworks under attack and (i)/(j) restricting and filtering connections — which can contain internal proxy traffic once underway, but the clause is primarily preventive and does not directly eradicate the proxy capability on compromised systems, leaving the response a slice of the full containment-and-eradication set.
- T1090.002partialdetects — A.8.20(f) requires logging and monitoring to enable recording and detection of actions affecting information security, which can surface anomalous outbound connections to an external proxy, but the control's scope is the organization's own network and it cannot observe the proxy-to-C2 leg or the proxy's own infrastructure.
- T1090.002partialprevents — A.8.20's network controls — particularly egress filtering (i), firewall rules, and authentication of systems on the network (h) — can block outbound connections to known or non-approved proxy destinations, preventing some uses of external proxies for C2; however, the proxy infrastructure is entirely outside the organization's control, the technique is designed to ride over legitimate-looking traffic paths, and the clause sets requirements rather than mandating strict egress filtering, so a substantial slice of the technique remains reachable.
- T1090.002partialresponds — A.8.20 has explicit response capabilities — item (m) drawbridges to isolate subnetworks 'if the network is under attack', item (i) firewalls to block the external proxy connection, item (n) disabling vulnerable protocols — which are real containment actions against ongoing C2 traffic. But A.8.20 is a network security control, not an incident response procedure: it does not assess whether the event is an incident, collect evidence, communicate, or close the incident, so it performs a slice of the response task set rather than the bulk of it.
- T1090.003partialdetects — A.8.20's monitoring (item f) and connection detection (item j) can surface traffic from known proxy exit nodes, anomalous routing patterns, and unauthorized system connections, but the technique is specifically designed to evade detection by chaining proxies so only the last hop is visible to the defender, leaving the chain itself unobservable.
- T1090.003partialprevents — A.8.20's filtering, firewall and authentication controls (items e, h, i, j) can block some multi-hop proxy traffic from entering the network, but the technique is specifically designed to evade source-based blocking — Tor has thousands of exit nodes, ORB networks use diverse compromised infrastructure, and new proxies are constantly added — so what is prevented is a slice rather than the bulk.
- T1090.003partialresponds — A.8.20 contains real response mechanisms — firewalls (i), connection filtering (j), and especially subnetwork isolation via drawbridges (m) — that can act on multi-hop proxy traffic once it is underway, but the technique's defining feature (chained proxies obscuring the source) is exactly what A.8.20 cannot reach: it can block the last hop and isolate affected segments, yet cannot trace or eradicate the upstream proxy chain, leaving the core obfuscation intact.
- T1090.004partialdetects — A.8.20's monitoring scope (item f) covers network traffic and can surface domain fronting through behavioral analysis of CDN traffic patterns or TLS inspection that compares SNI to HTTP Host, but the technique is specifically designed so the SNI shows a legitimate domain and the actual destination is hidden inside the encrypted tunnel, so standard network monitoring and firewall filtering (items i, j) do not catch it.
- T1090.004partialresponds — A.8.20 provides network-level containment mechanisms such as firewall filtering and subnetwork isolation that can sever the domain-fronted command and control channel, but it cannot eradicate the underlying malware from the compromised endpoint.
- T1095partialprevents — A.8.20 items (i) firewall filtering, (k) device hardening, (l) channel segregation, and especially (n) disabling vulnerable network protocols can prevent a slice of T1095 by removing unused non-application layer protocols and blocking others at the perimeter, but the technique's core protocols (ICMP, UDP) are required for network operation and cannot be disabled, and VMCI bypasses network-layer controls entirely, so the technique remains reachable.
- T1095partialresponds — A.8.20 has explicit response elements — (m) temporarily isolating critical subnetworks under attack and (n) disabling vulnerable network protocols — that can contain T1095's C2 channels once underway, but the control is primarily preventive and does not perform assessment, evidence collection, or foothold eradication, so it reaches only a slice of the response tasks.
- T1098partialdetects — A.8.20's logging and monitoring (item f) and network device hardening/logging can surface account manipulation on network devices (one of T1098's listed platforms), but the bulk of T1098 — Windows/AD, cloud IAM, identity provider, SaaS account changes — occurs at the identity layer outside A.8.20's network-scoped monitoring, so only a slice is detected.
- T1098.004partialdetects — A.8.20's network logging/monitoring (item f) and segregated administration channels (item l) can surface indicators such as SSH sessions from previously-unseen sources or management-channel commands like `ip ssh pubkey-chain` on network devices, but the core of T1098.004 is a host-level file modification that the network vantage point does not see, and cloud-API key insertions do not traverse the monitored network at all.
- T1102partialdetects — A.8.20(f) requires logging and monitoring of network actions, and (j) explicitly calls for detecting connections to the network, which can surface beaconing patterns, unusual data volumes, or connections to less-common web services used for C2; however, T1102 is specifically designed to blend with legitimate HTTPS traffic to common providers (Google, Microsoft, Twitter) using SSL/TLS, and the clause sets a monitoring requirement without mandating the deep inspection or behavioural baselining needed to reliably distinguish adversarial use from normal use, leaving a substantial undetected remainder.
- T1102partialprevents — A.8.20's egress filtering, network segmentation, and protocol restrictions can constrain T1102 by limiting which systems reach external services and which services are permitted, but the technique is specifically designed to use legitimate, commonly-allowed web services (Google, Microsoft, Twitter) over standard HTTPS, so network controls cannot block the traffic without breaking business functionality — the technique still runs, just over a narrowed set of paths.
- T1102partialresponds — The control includes temporarily isolating critical subnetworks during an attack, which contains the web service command-and-control traffic but does not eradicate the underlying compromise.
- T1102.001partialdetects — A.8.20(f) logging/monitoring and (j) detecting connections can surface some signals — unusual request frequency/timing to legitimate services, DNS lookups triggered when the dead drop resolver redirects to actual C2 infrastructure, and anomalous host behaviour — but the technique is engineered to blend into allowed, encrypted traffic to popular web services, so the embedded resolver content and the C2 handoff itself are not visible to network-layer monitoring.
- T1102.001partialprevents — A.8.20's network controls (egress filtering, protocol restrictions, disabling vulnerable protocols, restricting connections) can partially prevent dead drop resolver abuse by limiting which external services endpoints may reach, but the technique is specifically designed to use legitimate, commonly-allowed services (Google, Twitter) over encrypted HTTPS, so firewalls and network filtering cannot block it without breaking normal business traffic — the prevention is a slice, not the bulk.
- T1102.001partialresponds — A.8.20 includes temporarily isolating subnetworks when under attack, which contains the infected host's ability to reach the dead drop resolver, though it cannot eradicate the external resolver itself or the host infection.
- T1102.002partialdetects — A.8.20's network monitoring (f) and connection controls (i, j) can observe the C2 traffic since it traverses the organization's network, and behavioral analysis of beaconing patterns, data volumes, or known-bad indicators can surface some instances; however, the technique is specifically designed to blend with legitimate HTTPS traffic to common services using SSL/TLS, so encrypted content and normal-looking flows evade detection, and a conformant implementation focused on perimeter filtering rather than deep behavioral analysis would miss most instances.
- T1102.002partialprevents — A.8.20's firewall/egress filtering and network segmentation can restrict some outbound C2 channels, but T1102.002 is specifically designed to abuse legitimate, allowed web services (Google, Twitter, etc.) that cannot be blocked without breaking business operations, so the technique still runs even with the control in place.
- T1102.003partialdetects — A.8.20's network monitoring (item f) can record outbound HTTPS connections to legitimate services and observe beaconing patterns, but T1102.003 is specifically designed to blend with normal traffic to popular sites using SSL/TLS, so distinguishing C2 from legitimate use requires behavioral analysis or threat intelligence beyond what the clause mandates.
- T1102.003partialresponds — Isolating subnetworks or filtering connections during an attack severs the one-way command channel, providing network-level containment, but leaves the underlying compromised process running on the host.
- T1104partialdetects — A.8.20's item (f) requires logging and monitoring of network actions (referencing A.8.16/8.15), and item (j) requires detecting connections of equipment, so network telemetry can surface indicators of multi-stage C2 such as multiple distinct outbound channels, anomalous DNS, and fallback callbacks; however, T1104 is explicitly designed to obfuscate the C2 channel by hosting stages on separate infrastructure with no overlap, and A.8.20's monitoring is one item among many prevention-oriented controls rather than a dedicated detection mandate, so what is caught is a slice rather than the technique's full footprint.
- T1104partialresponds — A.8.20(m) explicitly provides response capability by isolating critical subnetworks (drawbridges) when under attack, and (i)/(j) allow firewall reconfiguration to block C2 channels; however, T1104 is specifically designed to survive single-channel blocking via fallback stages and backup callbacks, so containment reaches the contacted channel but not the campaign.
- T1105partialprevents — A.8.20's firewall filtering (i), protocol disabling (n), and network authentication (h) can block some ingress channels used for tool transfer, but T1105 has many vectors that traverse allowed protocols (HTTPS, DNS), legitimate web services, and cloud sync — channels network controls typically permit — and the technique presupposes an existing foothold, so what is prevented is a slice of the channels rather than the technique.
- T1105partialresponds — Item (m) of A.8.20 explicitly provides a response mechanism — temporarily isolating critical subnetworks via drawbridges 'if the network is under attack' — which is a genuine containment action against an in-progress ingress tool transfer. However, this is one containment mechanism scoped to critical subnetworks; the clause does not address eradication of the transferred tools, evidence collection, or the broader response lifecycle, so it covers a slice of `responds` rather than the bulk.
- T1110partialprevents — A.8.20 prevents a slice of brute force via network-level controls (firewall filtering of authentication services per (i)/(j), disabling vulnerable protocols per (n), network segmentation), but cannot prevent brute force against services that must remain reachable, and does not address offline brute force against stolen hashes at all — the bulk of prevention belongs to A.8.5 authentication controls.
- T1110.001partialdetects — A.8.20 item (f) explicitly calls for logging and monitoring to enable detection of security-relevant actions, and password guessing generates observable network traffic (repeated authentication attempts against services like SSH, RDP, HTTP), but detection is one item among fourteen and the clause's primary focus is prevention (firewalls, filtering, segregation, hardening); the actual detection depth is delegated to A.8.16/A.8.15, and network-level monitoring sees the attempts without necessarily distinguishing them from legitimate failed logins.
- T1110.001partialprevents — A.8.20 can prevent password guessing against services it shields at the network layer — firewalls filtering access to management ports (SSH, RDP, SMB, Telnet), disabling vulnerable protocols, and segregating networks — but it cannot prevent guessing against services that must remain reachable (externally-facing web apps, VPN endpoints, cloud SSO) and does not address authentication strength itself, which is A.8.5's domain.
- T1110.001partialresponds — A.8.20 has genuine response capability against password guessing once underway through (m) isolating critical subnetworks under attack (drawbridges), (i) firewall-based source blocking, and (j) detecting/restricting connections — but these are network-layer containment actions, not authentication-layer containment, and the clause's primary purpose is preventive; the actual stopping of the guessing requires authentication controls (A.8.5), so the response is a slice rather than the bulk.
- T1110.003partialprevents — A.8.20's network controls (firewalls, restricting/filtering connections, disabling vulnerable protocols, isolating subnetworks) can block external access to many of the targeted management services (RDP, SSH, SMB, Telnet, FTP, VNC) so password spraying against them never reaches the authentication step, but the technique explicitly targets services that must remain publicly reachable — SSO, cloud/federated applications, externally-facing email (e.g. Office 365) — which network controls cannot remove from the attack surface.
- T1110.003partialresponds — A.8.20 includes response-capable network actions — (m) temporarily isolating critical subnetworks under attack, (i) restricting/filtering connections, (j) detecting/restricting device connections — that can be applied against password spraying in progress (blocking source IPs, isolating affected segments), but it is not an incident response procedure: it does not assess, collect evidence, communicate, eradicate the credential compromise, or close the incident, so it performs a slice of the response task set rather than the whole.
- T1110.004partialdetects — A.8.20's clause (f) explicitly includes logging and monitoring to enable detection of actions relevant to information security, and credential stuffing generates observable network-level indicators (volume of authentication attempts, source patterns, connections to management services on common ports). However, the core detection of credential stuffing — authentication failures, account lockouts, credential reuse patterns — occurs at the authentication/application layer, which is A.8.16's domain. A.8.20 is primarily a network security control (firewalls, segregation, hardening) with detection as one aspect among many, so it catches a slice of the indicators rather than the bulk.
- T1110.004partialprevents — A.8.20's network-level controls (firewalls filtering/restricting connections per (i) and (j), disabling vulnerable protocols per (n), segmenting management services) can prevent credential stuffing against services that need not be internet-exposed (SSH, RDP, SMB, Telnet, FTP, etc.), but cannot prevent it against services that must remain publicly reachable (SSO, webmail, customer-facing apps, cloud federated auth) — those require authentication controls (A.8.5) to stop the technique.
- T1110.004partialresponds — A.8.20 item (m) explicitly provides a response action — temporarily isolating critical subnetworks 'if the network is under attack' — which can contain credential stuffing by blocking the source, but the bulk of the response to this technique (revoking compromised credentials, forcing resets, locking accounts) lives in A.8.5 and A.5.26, not in network security controls, so what A.8.20 contributes is a slice of the response set.
- T1111partialdetects — A.8.20's network logging and monitoring (item f) can surface some manifestations of T1111 — anomalous authentication proxying traffic, unexpected source patterns during token-based auth — but the technique's core mechanisms (endpoint keyloggers capturing smart card or RSA SecurID PINs, SMS interception, SMS-service compromise) occur at the endpoint or on external telecom infrastructure outside the network monitoring vantage point, so only a slice is observable.
- T1114partialdetects — A.8.20's network logging and monitoring (item f) and connection detection (item j) can surface some email collection — unusual access patterns to mail servers, large transfers to/from mail infrastructure, anomalous connections — but significant T1114 vectors fall outside network-level visibility: local client/PST access, cloud email API calls indistinguishable from legitimate user activity, and server-side forwarding rules that generate no anomalous traffic.
- T1114.002partialdetects — A.8.20(f) requires logging and monitoring of network actions, which can surface anomalous email-retrieval patterns (volume, source, timing) against on-premises Exchange, but the technique uses valid credentials over standard encrypted protocols and cloud-based collection (O365, Google Workspace) traverses only TLS to the provider — making most instances indistinguishable from legitimate mail traffic at the network vantage point.
- T1119partialdetects — A.8.20's network logging and monitoring (item f) can surface some artifacts of automated collection — unusual data flows, cloud API call patterns, lateral movement between systems — but the core of T1119 is host-based activity (reading files, querying databases, running scripts locally) that network-layer controls fundamentally cannot see, and cloud API collection often goes directly to provider endpoints outside the monitored path.
- T1127.002partialdetects — A.8.20's network monitoring (item f) and connection filtering (items i, j) can surface the download of .appref-ms/.application files and traffic to suspicious hosting infrastructure, but the technique's core mechanism — DFSVC.EXE launching the application locally — is endpoint execution outside A.8.20's network scope, leaving the bulk of the technique unobserved by this control.
- T1129partialdetects — A.8.20's monitoring scope is network traffic (per (f) and (j)), which can observe the UNC-path DLL loading variant of T1129 via SMB traffic to remote shares, but the bulk of T1129 — local module loading via LoadLibrary, dlopen, dlsym from on-disk paths — is host process activity outside any network vantage point the clause mandates.
- T1132partialdetects — A.8.20's logging and monitoring (clause f) and connection filtering (i, j) can surface some encoded C2 through traffic analysis and anomaly detection, but T1132 is specifically designed to make C2 content opaque to network-level inspection — Base64/gzip over standard protocols looks like normal traffic — and A.8.20 does not mandate the deep packet inspection or behavioral baselining needed to reliably distinguish encoded C2 from legitimate encoded data.
- T1132partialresponds — While isolating critical subnetworks under attack contains the encoded command and control traffic, it does not eradicate the host-side malware performing the encoding.
- T1132.001partialresponds — A.8.20 provides network-level response mechanisms — firewalls (i), temporary isolation of subnetworks (m), filtering — that can contain C2 traffic once identified, but the encoding technique itself is not eradicated (the adversary retains the capability and can re-encode or switch channels), and A.8.20 does not detect the encoding in the first place, so the response is a slice of containment rather than a full response to the technique.
- T1132.002partialdetects — A.8.20's logging and monitoring (item f) and network filtering (items i, j) provide vantage on C2 traffic flows and can surface statistical or protocol anomalies, but T1132.002 is specifically designed to make encoded C2 content look like normal protocol traffic, so the encoding itself evades the content-based detection the technique targets; what is caught is a slice of behavioral indicators rather than the encoding itself.
- T1132.002partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which contains the encoded command and control traffic but does not eradicate the malware performing the encoding.
- T1133partialdetects — A.8.20(f) requires logging and monitoring to enable recording and detection of actions affecting information security, which covers anomalous connections to external remote services, but the clause delegates scope to A.8.16 and A.8.15 rather than mandating the instrumentation depth that would surface every T1133 access attempt.
- T1133partialprevents — A.8.20 requires network access controls including authentication, filtering, and disabling vulnerable protocols, which can block unauthorized external remote service connections, but the technique also relies on valid accounts and exposed services that network-level controls alone cannot fully eliminate.
- T1133partialresponds — A.8.20 is primarily a preventive network security control, but item (m) explicitly describes a response action — temporarily isolating critical subnetworks (drawbridges) when the network is under attack — which is a genuine containment measure applicable to an adversary leveraging external remote services. The remainder of the clause (firewalls, filtering, hardening, authentication) is preventive, and the response is a slice of what full incident response entails.
- T1135partialdetects — A.8.20(f) requires logging and monitoring of network actions for detection, and T1135's SMB enumeration generates network traffic within that scope; however, the clause sets requirements rather than mandating instrumentation depth, so detection depends on implementation choices and some share-discovery queries blend with legitimate SMB traffic.
- T1135partialprevents — A.8.20's network segmentation (item m), filtering (item i), and protocol restrictions (item n) can limit the scope of share discovery across segments, but the technique still runs on any system where the adversary has a foothold within a permitted segment, and SMB cannot realistically be disabled entirely on Windows networks.
- T1136.001partialdetects — A.8.20's logging and monitoring of network devices (per (f) and (k)) can detect local account creation on network devices such as firewalls, routers and switches — a portion explicitly named in T1136.001 — but the bulk of the technique (account creation on Windows, Linux, macOS, ESXi and Kubernetes hosts) is a host-level event outside the network-security vantage point, so what is caught is a slice rather than most of the technique.
- T1136.001partialprevents — A.8.20 partially prevents T1136.001 through network device hardening (item k), network access controls and authentication (items h, i, j), which restrict unauthorized account creation on network infrastructure and firewall management consoles explicitly named in the technique; however, the bulk of the technique — local account creation on Windows, Linux, macOS, ESXi, and Kubernetes via commands like `net user /add` or `useradd` — runs on systems the adversary has already compromised, and A.8.20's network-layer controls do not reach command execution on those hosts.
- T1176partialdetects — A.8.20's network logging and monitoring (item f) can detect network-level manifestations of malicious extensions — C2 traffic, data tunneling, and exfiltration via extension functionality — but the core of T1176 (extension installation, in-application execution, and access to host application data) is an endpoint-level event outside A.8.20's network scope, so only a slice of the technique is observable.
- T1176.002partialdetects — A.8.20's network-layer monitoring (item f) and connection controls (items i, j) can surface some downstream network effects of a malicious IDE extension — unusual outbound connections from IDE processes, SSH tunnels, exfiltration traffic — but the core technique (extension installation and local code execution) is a host event outside the network control's vantage point, so only a slice is caught.
- T1185partialdetects — A.8.20's network logging/monitoring (item f) and connection detection (item j) can surface some indicators of T1185 — anomalous proxy connections, unusual pivots into internal resources, or traffic patterns inconsistent with normal user behaviour — but the core mechanism (browser-level injection that inherits cookies, sessions and certificates) produces traffic that is indistinguishable from the legitimate user's authenticated session, so the bulk of the technique remains invisible to network-layer controls.
- T1187partialdetects — A.8.20 requires logging and monitoring of network actions, which can surface forced SMB/WebDAV authentication traffic to external servers, but detection depends on the organization's chosen monitoring scope and the technique's use of standard protocols may blend with legitimate traffic.
- T1187partialprevents — A.8.20 can disable vulnerable network protocols (n) and restrict/filter connections (i), which can block SMB/WebDAV traffic to external adversary servers, but the technique can also be triggered through local files or internal EFSRPC abuse that network-level controls do not prevent.
- T1187partialresponds — A.8.20 includes explicit response capabilities — item (m) network isolation via drawbridges when under attack, and item (i) firewall-based connection restriction — that can contain an ongoing forced-authentication event by blocking outbound SMB to untrusted servers and isolating affected systems; however, the clause is primarily a preventive/detective network-security control, not an incident-response procedure, and it does not address eradication of already-exfiltrated credential hashes, leaving the response as a slice of the full verb.
- T1189partialdetects — A.8.20's network monitoring (clause f) and connection filtering (clauses i, j) can see the outbound connection to the malicious site, the download of exploit/script content, and any callback traffic, but the actual exploitation happens inside the browser on the endpoint — a layer A.8.20 does not cover. The network surfaces the traffic envelope; the in-browser exploitation and post-exploitation activity on the host are not visible to it.
- T1189partialprevents — A.8.20's network filtering (item i), protocol restrictions (n), and content controls (e) can block traffic to known malicious sites and filter exploit payloads, but drive-by compromise exploits the browser on the client side and frequently uses newly registered or legitimate-but-compromised sites that network controls cannot anticipate, so prevention is a slice rather than a closure.
- T1189partialresponds — A.8.20's drawbridge isolation (m), firewall filtering of C2 (i), and connection monitoring (j) provide real containment and eradication actions once a drive-by has executed — blocking outbound callbacks and isolating the compromised host from the rest of the network — but the technique's primary impact (browser exploit yielding code execution on the endpoint) is not undone by network-layer controls, and the response is bounded to network-side actions rather than the full response set.
- T1190partialprevents — A.8.20 prevents some instances of T1190 via firewalls (i), disabling vulnerable network protocols (n), hardening network devices (k), and network segmentation (m), but the technique specifically targets application-layer vulnerabilities in services that must remain Internet-reachable, and network controls do not address the application bug, glitch, or misconfiguration itself — the bulk of the technique's surface is untouched.
- T1190partialresponds — A.8.20 includes explicit response capabilities — item (m) temporarily isolating critical subnetworks under attack, item (j) restricting connections, item (i) filtering traffic — which are real containment actions against an exploited public-facing host. However, A.8.20 is a network-layer control: it can contain spread and cut off C2, but cannot eradicate the host-level foothold, patch the exploited vulnerability, perform forensics, or assess the incident. The response is a slice of the full response set, not the bulk with a bounded remainder.
- T1197partialdetects — A.8.20's network logging and monitoring (item f) can observe the file-transfer traffic BITS jobs generate, but BITS traffic is typically permitted through firewalls and resembles legitimate updater/messenger traffic, so distinguishing abuse at the network layer is hard; the BITS job creation, persistence configuration, and execution-on-completion semantics are host-level and outside A.8.20's network vantage point.
- T1199partialdetects — A.8.20's logging and monitoring (item f) and connection authentication (item j) can observe third-party traffic traversing the network, but T1199 is specifically designed to evade scrutiny by abusing legitimate trusted access paths that 'receive less scrutiny than standard mechanisms' — what is detected is a slice of anomalous behavior from partner connections, not the technique's core abuse of the trust relationship itself.
- T1199partialprevents — A.8.20's segmentation, filtering, and authentication of network connections (items i, j, l, m) can constrain where a compromised third party's access reaches, but the technique still executes through an authorized channel — the trust relationship itself is outside the clause's scope, so what is prevented is the blast radius, not the abuse.
- T1199partialresponds — A.8.20 contains explicit response mechanisms — (m) temporarily isolating critical subnetworks via drawbridges, (i) restricting/filtering connections, (j) detecting/restricting/authenticating device connections — that can contain an active T1199 intrusion by cutting off the third-party's access path or isolating affected segments. However, the technique exploits legitimate, authorized access (valid accounts from a trusted provider), so the control's response is bounded: isolating the third party disrupts legitimate business, filtering cannot easily distinguish malicious from authorized use of the same tunnel, and A.8.20 does not address credential eradication at the third-party provider. The response is real but partial.
- T1200partialdetects — A.8.20(j) explicitly requires detecting the connection of equipment and devices to the network, which catches the network-connected subset of T1200 (rogue wireless APs, network taps, MitM devices), but the technique also includes non-network hardware such as USB keystroke injectors and DMA attackers that never traverse the monitored network surface.
- T1200partialresponds — The control provides technical containment capabilities, such as isolating subnetworks or restricting connections, which can contain the network activity of a rogue device, but it cannot physically eradicate the hardware or contain local non-networked accessories.
- T1203partialdetects — A.8.20's network logging/monitoring (item f) and connection detection (item j) can surface indicators associated with client exploitation — malicious payload downloads, C2 callbacks, connections to known-malicious infrastructure — but the exploitation itself (memory corruption, code execution within the browser/Office/third-party app) is an application-layer event that network-layer controls do not directly observe.
- T1204.001partialdetects — A.8.20's network monitoring (f) and filtering (i, j) can detect the outbound connection to a malicious URL — DNS resolution, the HTTP(S) request, and any subsequent C2 or payload retrieval — but the core of T1204.001 is the user click and the resulting local execution, neither of which is visible at the network layer, and encrypted traffic further limits what the network sees.
- T1204.001partialprevents — Network filtering (firewalls, DNS/URL filtering per items i, j, n) can block traffic to known malicious destinations after a click, preventing the follow-on execution, but cannot stop the user from clicking and cannot reach destinations on legitimate-but-compromised sites, newly registered domains, or services that evade filtering — the technique is fundamentally user-driven and network controls are a secondary layer, not the primary mechanism.
- T1204.003partialdetects — A.8.20's network monitoring (item f) and connection controls (item j) can surface downstream indicators — suspicious image downloads, outbound C2 from a deployed instance, anomalous traffic from a freshly-spun workload — but the technique's core event (a backdoored image executing inside a VM or container) is a workload-level occurrence that network-layer monitoring does not directly observe; what is caught is a slice of network-visible side effects, not the technique itself.
- T1205partialdetects — A.8.20's logging and monitoring requirements (item f) and explicit detection of network connections (item j) provide a framework that can capture the signaling packets traversing the network — port-knock sequences, magic packets, and crafted connection attempts are observable in network telemetry — but the technique is specifically designed to look innocuous (failed connections to closed ports, magic values hidden in normal-looking traffic), and the control sets scope rather than mandating detection rules for traffic-signaling patterns, so what is caught is a slice the implementer chose to instrument.
- T1205partialprevents — A.8.20's controls (firewalls restricting connections, hardening of network devices, disabling vulnerable protocols, segregating administration channels) can prevent some traffic-signaling variants — particularly the network-device authentication bypass and signaling over vulnerable protocols — but the technique is designed to ride on traffic that looks normal, Wake-on-LAN operates below typical firewall control, and the technique presupposes a foothold A.8.20 does not address, so a bounded slice is prevented rather than the technique as a whole.
- T1205.001partialresponds — A.8.20(m) — temporarily isolating critical subnetworks if the network is under attack — is a genuine containment action against port knocking once the signal sequence is observed, and (i)/(j) filtering and authentication can be tightened in response; however, the clause does not address eradication of the port-knocking mechanism itself (the host-based firewall rule or custom software that opens the port), so containment is real but incomplete.
- T1205.002partialprevents — A.8.20's network segmentation, protocol restrictions, and filtering (items i, m, n) can limit which networks reach a host and thus constrain the crafted packet that triggers the backdoor, but the core of T1205.002 — installing a socket filter on a host with elevated privileges using libpcap/Winpcap — is a host-level action outside A.8.20's network-layer scope, so the filter still gets installed and can be triggered by traffic that does traverse permitted paths.
- T1207partialdetects — A.8.20 item (j) (detecting connection of equipment/devices to the network) and item (i) (firewall filtering) can surface a new machine joining the network to host a rogue DC, but the technique is explicitly designed to bypass the logging and SIEM monitoring that A.8.20 item (f) relies on, can reuse inactive registrations without a new device connection, and the core manipulation occurs at the AD schema level — outside the network-security vantage point A.8.20 occupies.
- T1210partialprevents — Network segmentation, device hardening, and protocol restrictions limit opportunities to exploit remote services for lateral movement.
- T1210partialresponds — A.8.20 item (m) explicitly addresses containment by temporarily isolating critical subnetworks (e.g., with drawbridges) when the network is under attack, which is a genuine response action against lateral movement; however, the control is primarily preventive (firewalls, filtering, hardening, disabling vulnerable protocols) and the response capability is limited to network isolation, not the broader response tasks of eradication, evidence collection, or closure.
- T1211partialdetects — A.8.20's network-level logging/monitoring (item f) and connection detection (item j) can surface some network artifacts of exploitation attempts, but T1211 is specifically designed to evade detection by suppressing logging, hiding activity, and operating within trusted or unmonitored components — and much of the technique (exploiting logging pipelines, security tools, cloud infrastructure) operates outside A.8.20's network-level scope.
- T1211partialprevents — A.8.20 prevents a slice of T1211 through hardening of network devices (k), disabling vulnerable network protocols (n), network segregation (l, m), and connection filtering (i, j), which close off exploitation of network-layer vulnerabilities; the remainder is the bulk of the technique, which targets application, cloud, SaaS, and logging-pipeline vulnerabilities outside the network-security control's scope.
- T1212partialdetects — A.8.20 item (f) explicitly mandates network logging and monitoring for detection, which catches network-observable forms of T1212 (replay attacks, network-borne exploitation attempts, anomalous authentication traffic), but a substantial portion of the technique — local kernel/OS exploitation, on-host credential store attacks, and cloud identity-provider exploitation — occurs outside the network vantage point the control monitors.
- T1212partialprevents — A.8.20 prevents a slice of T1212 through disabling vulnerable network protocols (item n), network segmentation, firewalls, and authentication of network connections, but the clause addresses network-layer protections rather than the vulnerability management and patching that T1212 fundamentally depends on; exploitation of OS, application, and identity-provider vulnerabilities — including local exploits and cloud token forgery — falls outside the network-security scope.
- T1212partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which provides network-level containment of remote exploitation attempts, though it does not eradicate the vulnerability or address local exploitation.
- T1213partialdetects — A.8.20(f) logs and monitors network actions, which can surface anomalous data flows from repository services, access from unusual locations, or unusual connection patterns to SharePoint/Confluence/databases; however, T1213's core mechanism is application-layer reading of repository data using legitimate credentials and authorized access paths, which network-layer monitoring largely cannot distinguish from normal use.
- T1213partialprevents — A.8.20 restricts and filters system connections, disables vulnerable protocols, and hardens network devices, which reduces the attack surface an adversary can use to reach an information repository, but it does not control access to the repository itself once a connection is established, leaving a substantial remainder.
- T1213.003partialprevents — A.8.20's network segmentation, firewalling and authentication of systems on the network can restrict who reaches internal code repositories, but the technique's core mechanism is application-layer data collection from a repository (often third-party-hosted, where A.8.20 has no reach), and once network access is gained the read happens at the repository service itself — so network controls address only the access path, not the collection.
- T1213.006partialdetects — A.8.20 item (f) requires logging and monitoring of network actions relevant to information security, and database queries traverse the network, so traffic to/from database servers is observable — anomalous query volumes, large data transfers, and unexpected connection sources can surface mining activity. However, the technique is fundamentally about data access at the database layer, and A.8.20's monitoring scope is the network transport rather than database activity; distinguishing malicious queries from legitimate ones typically requires database-level auditing or user behavior analytics outside A.8.20's scope, so detection is a slice rather than a bounded remainder of a working mechanism.
- T1213.006partialprevents — A.8.20's network controls (segmentation, firewalls, system authentication, restricting/filtering connections) can block direct network paths to database servers and prevent lateral movement to database segments, but the technique still executes through legitimate application channels, compromised application servers with database access, cloud-hosted database APIs over web protocols, and credentialed access — none of which network-layer controls reach.
- T1213.006partialresponds — A.8.20's drawbridge capability (m) and network filtering (i) can isolate the database's network segment and block the adversary's network access once database mining is observed, which is a genuine containment action; however, this addresses only the network-borne slice of the response — the data already mined is not undone, database-level access and persistence are outside A.8.20's scope, and the adversary may retain non-network access paths.
- T1216.001partialdetects — A.8.20's network logging and monitoring (item f) can observe the outbound HTTP(S) request PubPrn makes to fetch the remote .sct scriptlet, which is a necessary component of the technique; however, the technique's primary mechanism is host-side execution via cscript.exe and pubprn.vbs, which is outside A.8.20's network-scoped monitoring surface, so only a slice of the technique's observable behavior is caught.
- T1218partialdetects — A.8.20's network monitoring (item f) can detect network-side artifacts of T1218 such as payload downloads and C2 traffic from the proxy binary, but the core technique — proxying execution through a trusted local binary — is an endpoint event that network monitoring does not see.
- T1218.005partialdetects — A.8.20's network monitoring (clause f) and connection filtering (clauses i, j) can observe the outbound HTTP/HTTPS traffic mshta generates when fetching remote .hta files or scripts, but the technique's core mechanism — mshta.exe executing HTA content on the endpoint — is outside the clause's network scope, so only the network-side slice is caught.
- T1218.007partialdetects — A.8.20's network monitoring (item f) and connection filtering (items i, j) can detect network-level indicators of msiexec abuse such as remote MSI file retrieval over SMB/HTTP and subsequent C2 traffic, but the core of T1218.007 is local host-based process execution — command-line arguments, parent-child process relationships, and DLL execution — which falls outside A.8.20's network-scoped monitoring mandate.
- T1218.010partialdetects — A.8.20's monitoring scope is network-level (traffic, connections, network devices per (f) and (j)), so it can detect the network-aware Squiblydoo variant's URL fetch and remote COM scriptlet loading, but the core technique — regsvr32.exe execution, command-line arguments, and module loading — is an endpoint activity outside A.8.20's scope.
- T1219partialprevents — A.8.20's network controls (firewall filtering, system authentication, network segmentation, disabling vulnerable protocols, restricting connections) can block specific remote access tools, constrain outbound C2 channels, and limit which systems accept remote sessions, but the technique deliberately abuses legitimate tools that organizations commonly allow for business reasons, so what is prevented is a slice rather than the bulk.
- T1219partialresponds — A.8.20 can contain a remote access tool session once underway via item (m) network isolation/drawbridges and item (i) firewall blocking of C2 traffic, but it does not eradicate the tool itself — removal of the installed software and persistence requires A.8.7 or endpoint controls, so the response is limited to network-level containment.
- T1219.001partialdetects — A.8.20 has a network vantage point and its logging/monitoring (item f) and connection-detection (item j) clauses can observe the SSH/tunnel sessions IDE tunneling generates, but the technique is explicitly designed to blend in with legitimate developer workflows using encrypted SSH and legitimate account authentication (e.g., GitHub), so what is reliably detected is a slice of anomalous patterns rather than the technique itself.
- T1219.001partialprevents — A.8.20's network filtering, protocol restrictions, and egress controls can constrain IDE tunneling by blocking outbound SSH or proprietary tunnel protocols from non-developer systems, but the technique uses legitimate protocols and authentication (e.g., GitHub) that developers require, so it cannot be fully prevented without breaking legitimate development workflows.
- T1219.001partialresponds — A.8.20 provides the capability to temporarily isolate subnetworks or filter connections when under attack, which contains the technique by severing the IDE tunnel's command and control channel, but it cannot eradicate the host-level IDE configuration or persistence mechanisms.
- T1219.002partialprevents — A.8.20's network filtering, connection restriction, system authentication, and protocol disabling can block outbound connections to remote desktop service providers and prevent unauthorized instances, but the technique specifically uses legitimate software (TeamViewer, VNC, AnyDesk, etc.) that organizations commonly allow for support purposes, so network controls cannot fully prevent the technique where legitimate use is required.
- T1219.003partialdetects — A.8.20's items (f) logging/monitoring and (j) detecting the connection of equipment and devices to the network apply to a hardware device that must join the network and generate traffic, but the technique is explicitly designed to blend in as legitimate remote-access hardware allowed by peripheral device policies and to bypass software security solutions, so what is detected is the device's presence and traffic patterns rather than its malicious use.
- T1219.003partialprevents — A.8.20's network filtering, authentication of systems on the network, and restriction of equipment connections (items h, i, j) can prevent unauthorized network access from rogue remote-access hardware, but T1219.003 specifically uses legitimate KVM-class hardware that is commonly allowed by peripheral device policies and is noted to bypass software security solutions, so a meaningful slice of the technique survives network controls.
- T1219.003partialresponds — A.8.20's guidance items (m) isolating critical subnetworks via drawbridges, (j) restricting connections of equipment, and (i) firewall filtering are genuine containment actions against a hardware-based C2 channel already in use; however, the technique involves physically installed hardware whose full eradication requires physical removal outside A.8.20's scope, so the response is bounded to network-layer containment and traffic阻断 rather than complete removal of the foothold.
- T1220partialdetects — A.8.20's network logging and monitoring (clause f) can surface the network-side signals of T1220 — outbound fetches of remote XSL files, DNS lookups for attacker-controlled hosts, and traffic from msxsl.exe/wmic — but the technique's core is local script execution via XSL processing, which network monitoring does not see.
- T1221partialdetects — A.8.20's network monitoring and logging (clause f) can observe the outbound URL fetch and any forced authentication attempt that template injection triggers, but the core of T1221 — modifying template references inside OOXML/RTF documents — happens at the application/document layer and is not visible to network controls; what is detected is the delivery side-effect, not the technique itself.
- T1485partialdetects — A.8.20's item (f) logging/monitoring and item (m) subnetwork isolation can surface network-level indicators of T1485's worm-like propagation and large-scale SMB/admin-share activity, but the core data destruction (file overwrite/deletion) is a host- and storage-level event outside the clause's network-focused scope, so only a slice is detected.
- T1485partialresponds — A.8.20's drawbridge isolation (m), firewall filtering (i), and connection restriction (j) are direct response actions against the worm-like network propagation T1485 explicitly describes, but the control does not address the core data-destruction act on individual systems nor restore overwritten data, so it engages a slice of the technique's spread vector rather than the technique as a whole.
- T1489partialdetects — A.8.20's network-focused logging and monitoring (clause f) can detect service stops that manifest as loss of network-reachable services or that traverse the network (remote admin, cloud API calls like DisableAPIServiceAccess), but cannot detect local service stops of non-network services or process kills that produce no network-visible artifact, which is a substantial share of T1489's usage.
- T1491.002partialdetects — A.8.20's network logging and monitoring (item f) can detect network-level indicators associated with an external defacement attack — exploit traffic, anomalous connections, post-defacement traffic shifts — but the defacement itself is an application/web-server content change, and the clause's scope is network-layer protection (segmentation, filtering, device hardening) rather than content or file-integrity monitoring, so what is caught is a slice of precursors and side-effects rather than the event itself.
- T1491.002partialprevents — A.8.20's network-layer controls (firewalls, segmentation, protocol restrictions, authentication of network connections, hardening of network devices) can prevent some defacement vectors by restricting access to web server management interfaces and limiting lateral movement, but the primary vectors for external defacement are web application and CMS vulnerabilities, which sit above the network layer and are not addressed by this clause.
- T1495partialprevents — A.8.20's network controls (firewalls/filtering, authentication of systems, hardening of network devices, disabling vulnerable protocols) can prevent some vectors for firmware corruption—particularly on network devices themselves, which the technique explicitly targets—but cannot prevent firmware corruption that occurs after an adversary gains system-level access through other means, nor does it address firmware integrity on non-network devices (motherboards, hard drives, video cards).
- T1496partialdetects — A.8.20(f) requires logging and monitoring of network actions, which can surface the anomalous traffic patterns of resource hijacking (e.g., cryptomining pool connections or proxyjacking), but the control's scope is network-level and will miss hijacking that stays within a host's compute resources without generating distinct network signatures.
- T1496partialprevents — A.8.20's network-layer controls (firewalls filtering outbound connections to mining pools/proxy networks, segmentation limiting lateral spread, disabling vulnerable protocols, NAC restricting unauthorized devices) prevent some resource-hijacking vectors, but the technique commonly runs on already-compromised hosts using legitimate protocols or via cloud credential abuse, which network controls do not reach.
- T1496partialresponds — A.8.20 has genuine response mechanisms against resource hijacking — (m) network isolation via drawbridges, (i) firewall filtering of outbound C2/mining/proxy traffic, (n) disabling abused protocols — but these act only at the network layer; host-level compute abuse (CPU mining, local resource consumption) continues even when the system is isolated, and the control does not eradicate the underlying compromise.
- T1496.001partialdetects — A.8.20's network logging and monitoring (item f) and connection detection (item j) can surface network-level indicators of compute hijacking — outbound connections to mining pools, DNS queries to mining domains, and unusual traffic patterns — but the technique's primary signature is host-level resource consumption (CPU/GPU spikes, process execution, container deployment), which sits outside the clause's network-focused scope.
- T1496.002partialdetects — A.8.20 explicitly requires logging/monitoring of network actions (item f, cross-referencing 8.16) plus detection-oriented items (j: detecting device connections; m: isolating subnetworks under attack; n: disabling vulnerable protocols), which surface bandwidth anomalies, proxyjacking traffic, scanning, or botnet C2 on the monitored estate; the bounded remainder is pre-compromise external scanning or hijacking of unmonitored/off-premises bandwidth.
- T1496.002partialprevents — A.8.20's network controls — egress filtering, connection restrictions, system authentication, and the ability to temporarily isolate subnetworks — directly constrain a compromised system's ability to use its bandwidth for botnet, proxyjacking, scanning, or torrent activity, which is the network-facing half of T1496.002. But the technique presupposes endpoint compromise, which A.8.20 does not address, and some bandwidth-hijacking vectors (local resource exhaustion, internal scanning) are not reachable by network-perimeter controls alone, so what is prevented is a slice rather than the bulk.
- T1496.002partialresponds — A.8.20's clauses (m) isolating critical subnetworks under attack, (i) restricting/filtering connections, and (j) detecting/restricting device connections provide genuine network-level containment against bandwidth hijacking in flight — isolating the compromised systems or filtering their outbound traffic stops the abuse. But the clause does not identify which endpoints are compromised (that is A.8.16's job) and does not eradicate the underlying malware, so it performs the containment slice of response without the identification or eradication slices.
- T1498partialdetects — A.8.20 explicitly requires logging/monitoring of actions affecting network security (item f, cross-referencing 8.16), hardening (k), restricting/filtering connections (i), detecting device connections (j), and isolating subnetworks under attack (m), all of which surface Network DoS traffic or its effects across the technique's scope on supported platforms.
- T1498partialprevents — A.8.20's filtering, disabling of vulnerable protocols, hardening, and drawbridge isolation prevent some DoS vectors (protocol amplification, reflection via spoofed sources, certain application-layer abuse), but cannot prevent volumetric DDoS that exhausts the link bandwidth itself — the same component-level bandwidth limit that bounds A.8.14's `recovers` against this technique bounds `prevents` here.
- T1498partialresponds — A.8.20 includes explicit response actions against a network attack in progress — item (m) on temporarily isolating critical subnetworks via drawbridges, and item (i) on firewalls that can filter attack sources — but the clause is primarily a preventive posture-setting control, and the response actions are a subset of its scope rather than a complete incident-response procedure (assessment, evidence collection, eradication belong to A.5.25/A.5.26/A.5.28).
- T1498.001partialprevents — A.8.20 includes preventive elements — firewalls/filtering (i), disabling vulnerable protocols (n), authenticating systems on the network (h), and restricting connection of equipment (j) — that block some flood vectors such as protocol abuse and traffic from unauthorized sources, but the clause itself states additional controls are required for availability, and a volumetric DDoS from a distributed botnet still reaches the network perimeter where these measures cannot stop it.
- T1498.001partialresponds — A.8.20 includes explicit response mechanisms for an attack in progress — item (m) drawbridges to isolate critical subnetworks, item (i) firewall filtering to drop attack traffic, item (j) restricting connections — but these do not fully contain a direct volumetric flood: the flood still saturates the network link at the perimeter, and distinguishing DDoS from legitimate traffic is exceedingly difficult per the technique description, so what is contained is the impact on internal systems rather than the flood itself.
- T1498.002partialdetects — A.8.20(f) requires logging and monitoring to enable recording and detection of actions affecting information security, which can surface the traffic anomaly of a reflection amplification attack, but the control's scope is set by business and security requirements rather than mandating the depth of instrumentation needed to distinguish this specific volumetric pattern from other network events.
- T1498.002partialprevents — A.8.20 can prevent the organization from contributing to the reflection ecosystem (item n: disabling vulnerable protocols like open DNS resolvers, NTP, memcached on its own services) and can filter some reflected traffic at the perimeter (items i, j), but the attack traffic originates from third-party reflectors the organization does not control, and the volumetric nature overwhelms any reasonable filtering, so the technique still runs and largely succeeds against the target.
- T1498.002partialresponds — A.8.20(m) drawbridges explicitly isolate critical subnetworks under attack and (i)/(j) filtering at firewalls can drop attack traffic once characterised, so real containment actions exist; however, reflection amplification uses legitimate protocols (DNS, NTP, memcache) from legitimate third-party servers, the volume typically saturates upstream capacity before filtering engages, and blocking reflectors wholesale is impractical, so the response reaches a slice of the attack rather than containing it.
- T1499partialdetects — A.8.20(f) requires logging and monitoring to enable recording and detection of actions affecting information security, which covers endpoint DoS events observable on the network, but the clause delegates scope to A.8.16 and A.8.15 rather than mandating the depth of instrumentation needed to surface application-layer exhaustion reliably.
- T1499partialprevents — A.8.20's network-layer controls (firewalls/filtering per (i), disabling vulnerable protocols per (n), hardening network devices per (k), isolating critical subnetworks per (m)) prevent some endpoint DoS vectors—particularly those exploiting known-vulnerable protocols or arriving from filterable sources—but the technique's core vectors (application-layer exhaustion, legitimate-looking botnet traffic, endpoint resource exploitation) reach the target through traffic network controls cannot meaningfully distinguish, so what is prevented is a slice rather than the bulk.
- T1499partialresponds — A.8.20 includes explicit response capabilities — (m) temporarily isolating critical subnetworks during attack, dynamic application of (i) filtering and rate limiting, and (j) restricting connections — that can contain an Endpoint DoS in progress by dropping malicious traffic before it exhausts endpoint resources. However, the clause is primarily preventive (establishing network security controls), the response actions are a subset of its scope, and application-layer exploits or resource exhaustion that bypass network-level controls are not reached, so the response is a slice rather than the bulk.
- T1499.001partialresponds — A.8.20 includes explicit response mechanisms — drawbridges (m) to isolate subnetworks under attack, real-time firewall filtering (i), and monitoring (f) — that act once the flood is underway, but the clause is primarily preventive (hardening, segregation, authentication, protocol disabling) and the response slice is bounded: network-layer controls can drop malicious packets and isolate segments, yet the OS-level exhaustion on the endpoint itself is not directly contained, and the attack continues from outside the perimeter.
- T1499.002partialprevents — A.8.20 includes preventive mechanisms relevant to service exhaustion floods — firewalls and connection filtering (i, j), disabling vulnerable protocols such as SSL renegotiation (n), and isolating subnetworks under attack (m) — but these cannot prevent the attack from being launched or from reaching the network edge, since HTTP floods use legitimate-looking requests that are hard to distinguish from real traffic and the volume-based variants overwhelm capacity regardless of filtering.
- T1499.002partialresponds — A.8.20 includes explicit response elements against an ongoing network attack — (m) temporarily isolating critical subnetworks with drawbridges when the network is under attack, and (i)/(j) restricting and filtering connections during the event — but the clause is primarily a preventive network-security control, and the response actions it names (isolation, filtering) bound the impact rather than eradicate the actor, which is the shape `partial` describes.
- T1499.003partialprevents — A.8.20 requires network segregation, filtering, and disabling vulnerable protocols, which can block or restrict the malicious traffic reaching the application, but the technique's exploitation of legitimate application features means it cannot be fully prevented at the network layer alone.
- T1499.003partialresponds — A.8.20 item (m) explicitly enables temporarily isolating critical subnetworks (drawbridges) if the network is under attack, and item (i) firewalls can be reactively reconfigured to filter attack traffic — both are containment actions once the technique is underway. However, T1499.003 exhausts application-layer resources, so network-level isolation and filtering contain the blast radius without addressing the resource exhaustion itself; the response reaches the network path, not the application mechanism being abused.
- T1499.004partialdetects — A.8.20's logging/monitoring (f), connection detection (j), and isolation under attack (m) can surface network-borne exploit attempts and the resulting service unavailability, but T1499.004's core mechanism — software vulnerability exploitation crashing applications/systems — happens at the application/host layer where A.8.20 has no instrumentation, leaving a named remainder of crashes invisible to network-layer detection.
- T1499.004partialprevents — A.8.20's network controls (segmentation, firewalls, disabling vulnerable protocols, filtering connections) can prevent some exploitation paths by limiting network exposure and restricting which systems reach vulnerable services, but the technique exploits application/system software vulnerabilities that network-layer controls cannot address when the service must remain reachable -- so prevention is a slice, not the bulk.
- T1499.004partialresponds — A.8.20 contains an ongoing exploitation via (m) isolating subnetworks under attack and (i) filtering/restricting connections, but it does not eradicate the underlying software vulnerability and cannot undo systems already crashed — containment is real, eradication is not, so the response is a slice of the verb.
- T1505partialdetects — A.8.20's network logging and monitoring (item f) can detect some downstream network indicators of a malicious server component (C2 traffic, unusual connections from compromised servers), but the core technique — installing malicious extensions/modules on server applications — is a host/server-side action outside the clause's network-security scope, so only a slice is caught.
- T1505partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which contains the adversary's network access to the compromised server but does not eradicate the malicious software component itself.
- T1505.002partialdetects — A.8.20's network monitoring (item f) can observe downstream effects of a malicious transport agent — anomalous email traffic patterns, unexpected outbound connections, or unusual SMTP behaviour — but the core technique (registering a .NET assembly as a transport agent on the Exchange server) is a server-side action outside the network-security control's primary vantage point, so what is detected is a slice of indicators rather than the technique itself.
- T1505.003partialdetects — A.8.20's network logging/monitoring (item f) and connection filtering (items i, j) can surface some web shell C2 traffic — anomalous outbound connections from web servers, beaconing patterns, traffic to known-bad destinations — but the core of T1505.003 (placement and execution of the script on the web server) is an application/file-level event outside the clause's network vantage point, and well-crafted web shell traffic resembles normal HTTP/HTTPS.
- T1505.003partialresponds — A.8.20's guidance includes (m) drawbridges for isolating subnetworks under attack and (i) firewall-based filtering, which are genuine containment actions against a web shell's C2 traffic and lateral movement; however, the clause does not remove the web shell from the web server, and its primary function is preventive, so the response is a slice of what full containment and eradication requires.
- T1505.004partialdetects — A.8.20's network monitoring (item f) and traffic inspection can surface the running effects of malicious IIS components — anomalous HTTP request/response modification, proxied C2, or unusual outbound patterns from the web server — but the installation itself (DLL drop, ISAPI/module registration) is a host-level filesystem and IIS-configuration event that network-layer controls do not see, leaving the core persistence action undetected.
- T1525partialdetects — A.8.20's network monitoring (clause f) and connection authentication (h, j) can surface peripheral indicators of T1525 such as anomalous API calls to a container/image registry or unusual traffic from a compromised instance to cloud control planes, but the core action — modifying image contents in a registry — is not a network event and requires image-integrity or registry-level monitoring that lies outside A.8.20's network-security scope.
- T1526partialdetects — A.8.20(f) covers network-level logging and monitoring, and cloud service discovery API calls (Microsoft Graph, Azure Resource Manager, etc.) traverse the organization's network to reach cloud endpoints, giving the network some vantage point. However, the technique is fundamentally about enumerating cloud-side resources via legitimate APIs, which is far more naturally surfaced by cloud-native telemetry (CloudTrail, Azure Activity Log) than by network monitoring; encrypted HTTPS payloads and the resemblance of discovery calls to normal cloud service usage limit what network-level detection can reliably distinguish.
- T1529partialdetects — A.8.20's logging and monitoring of network actions (item f) and authentication of network connections (items h, j) surface network-mediated shutdowns/reboots — remote `reload` commands on network devices, remote Windows API shutdowns, and hypervisor/cloud-console actions that traverse monitored links — but the control's scope is the network, not the endpoint, so locally-initiated shutdowns, BSODs via local API calls, and shutdowns executed entirely within a host's own execution context produce no network event for A.8.20 to observe.
- T1529partialprevents — A.8.20 prevents the network-reachable slice of T1529 — remote shutdown via Network Device CLI is blocked by firewalling (i), system authentication (h), device hardening (k), segregated admin channels (l), and disabled vulnerable protocols (n) — but the technique's bulk runs through local OS commands, Windows API calls (InitializeSystemShutdownExW, NtRaiseHardError), and hypervisor/cloud consoles, none of which network controls reach.
- T1530partialdetects — A.8.20's network monitoring (item f) and connection detection (item j) can surface some indicators — unusual outbound volumes to cloud APIs, connections from unauthorized devices, anomalous traffic patterns — but the core of T1530 (data access at the cloud provider via authenticated API calls, misconfigured buckets, leaked credentials) happens at the cloud provider's infrastructure, where the organization's network monitoring has limited visibility; legitimate-looking encrypted API traffic to external services is largely indistinguishable from normal cloud use without cloud-specific telemetry.
- T1534partialdetects — A.8.20's network monitoring (item f) and connection logging give a vantage point on internal traffic and could surface some indicators of an internal spearphishing campaign (anomalous volumes from a compromised account, callbacks to external infrastructure), but the core detection of the phishing content itself — the malicious link, attachment, or chat payload — is not what network-layer controls do, so what is caught is a slice rather than the technique.
- T1539partialdetects — A.8.20's network logging/monitoring (item f) and connection detection (item j) can surface some cookie theft via the AitM vector (e.g., Evilginx2/Muraena traffic to attacker infrastructure, anomalous proxy patterns), but the dominant vectors — local disk and process-memory theft — are outside network security's scope, and even the in-transit case requires specific indicators to distinguish from legitimate TLS to the real service.
- T1539partialprevents — A.8.20's network-layer controls (encryption in transit per (e), connection filtering/authentication per (h)-(j), disabling vulnerable protocols per (n)) prevent the network-sniffing vector of cookie theft, but the technique's other vectors — local disk/memory theft, malicious JavaScript injection, and AitM via tools like Evilginx2 that terminate TLS at a proxy — sit outside the clause's scope, so only a slice of the technique is prevented.
- T1542.004partialdetects — A.8.20's network monitoring (f) and connection detection (j) can surface the TFTP transfer or unexpected device reboots that accompany a ROMMONkit attempt, and device hardening (k) can include firmware integrity verification, but the attack is specifically designed to be difficult to detect, operates below the OS at the boot-loader level where on-device telemetry is untrusted once compromised, and the signals available to network-level monitoring are narrow.
- T1542.004partialprevents — A.8.20 requires disabling vulnerable network protocols and hardening network devices, which can prevent ROMMONkit by removing the TFTP or other management vectors it relies on, but the control is a set of requirements rather than a universal mechanism, so the technique remains reachable on devices where those protocols are still permitted or where the hardening is incomplete.
- T1542.004partialresponds — A.8.20(m) drawbridge isolation and (i) filtering/restricting can contain a compromised network device once the ROMMONkit is known, but the clause does not address eradication of the firmware implant itself — re-flashing ROMMON is outside its scope, so containment engages while eradication does not.
- T1542.005partialdetects — A.8.20(f) requires logging and monitoring of actions affecting information security, and (j) explicitly covers detecting connections of equipment and devices to the network; TFTP traffic and configuration changes to network devices fall within this monitoring scope. However, the clause sets a monitoring requirement rather than mandating specific instrumentation of network device boot processes or firmware integrity, so the actual image loading on the device is not necessarily surfaced — what is caught is a slice of the technique's observable footprint, not the bulk.
- T1542.005partialprevents — A.8.20's guidance on hardening network devices (k), authenticating systems on the network (h), restricting connections (i,j), disabling vulnerable protocols (n), and segregating administration channels (l) provides mechanisms that can be applied to constrain TFTP boot abuse, but the technique exploits a legitimate network administration function and the clause does not specifically mandate boot-image integrity verification, signed images, or protection of the TFTP server itself, so what is prevented is a slice rather than the bulk.
- T1542.005partialresponds — A.8.20 includes temporarily isolating critical subnetworks during an attack, which can contain a compromised network device or block the malicious TFTP transfer, but it does not eradicate the unauthorized boot configuration.
- T1543.005partialdetects — A.8.20's network logging and monitoring (item f) can surface network-level artifacts of container service activity — kubelet/Docker API calls, image pulls, and outbound connections from deployed containers — but the core technique operates at the host level (modifying daemon configurations, creating containers with restart=always, deploying DaemonSets), which is outside a network-security control's monitoring scope.
- T1550partialdetects — A.8.20's network monitoring and authentication controls (items f, h, i, j) can surface lateral movement patterns and anomalous connection behaviour associated with T1550, but the technique uses legitimately-issued hashes, tickets and tokens whose reuse is largely indistinguishable from valid authentication at the network layer, so what is detected is a slice of the movement rather than the use of the material itself.
- T1550partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which contains the adversary's ability to move laterally across the network using the stolen authentication material, though it does not eradicate the compromised tokens or hashes.
- T1550.002partialdetects — A.8.20(f) explicitly mandates logging and monitoring for detection of actions relevant to information security, and network-level monitoring can surface lateral movement patterns and anomalous authentication sources indicative of PtH; however, the core mechanism uses valid hashes so the authentication itself looks legitimate at the network layer, and the bulk of PtH detection (credential theft, LSASS access, NTLM-vs-Kerberos anomalies) belongs to host and authentication-log controls outside this clause's scope.
- T1550.002partialprevents — A.8.20's network segmentation (m), firewalling (i), protocol restrictions (n) and admin-channel segregation (l) constrain the lateral paths over which a stolen hash can be replayed, but the technique itself is an authentication-layer bypass that succeeds against any reachable system the network permits; A.8.20 does not address the authentication mechanism, so what is prevented is the scope of movement, not the technique.
- T1550.002partialresponds — The control includes temporarily isolating critical subnetworks during an attack, which contains lateral movement via Pass the Hash across network segments but does not eradicate the compromised credentials or stop intra-subnet movement.
- T1550.003partialprevents — A.8.20 requires network segmentation, device hardening, and disabling vulnerable protocols, which can block the lateral movement that pass-the-ticket enables, but it does not prevent the initial credential theft or ticket creation on a compromised host.
- T1552.005partialdetects — A.8.20's network monitoring (item f) and connection detection (item j) can surface some instances of T1552.005 — particularly SSRF exploitation where a web server's request to 169.254.169.254 traverses monitored network paths — but direct queries from within a cloud instance to the link-local metadata address typically do not traverse network monitoring points and remain invisible to A.8.20's scope.
- T1552.005partialprevents — A.8.20 requires network segmentation, firewalling, and disabling vulnerable protocols, which can block or restrict access to the cloud metadata API from compromised instances, but the API is a legitimate service and cannot be fully disabled without breaking cloud functionality.
- T1552.007partialdetects — A.8.20's network monitoring (item f) can log container API calls traversing monitored segments, but the technique's credential retrieval via Docker/Kubernetes APIs often occurs on east-west overlay traffic that may fall outside the monitored scope, and legitimate-looking API calls are not flagged as credential theft without specific signatures.
- T1552.007partialprevents — A.8.20's network controls (firewalls, segmentation, authentication of network connections, restricting/filtering system connections) can prevent external or unauthenticated access to the Docker/Kubernetes API endpoints, but the technique explicitly notes adversaries with a pod's service account or in-cluster foothold reach the API from inside the network where perimeter and segmentation controls do not apply.
- T1552.007partialresponds — A.8.20 includes temporarily isolating critical subnetworks during an attack, which provides network-level containment by severing remote access to the container API, though it does not eradicate the adversary's presence or revoke the gathered credentials.
- T1555partialdetects — A.8.20's network logging and monitoring (item f) can detect network-based slices of T1555 — access to cloud secrets vaults, AD queries, and exfiltration of credential stores over the network — but the core action of reading local password stores (browser credential DBs, OS keychains, local password managers) is host-resident and outside the clause's network vantage point.
- T1555.006partialdetects — A.8.20's network monitoring (item f) can observe traffic to cloud secrets manager endpoints and detect anomalous patterns of API calls, but the calls are TLS-encrypted so the network cannot see which secrets are being requested or whether the request is authorized — meaningful detection of T1555.006 requires cloud-level logging (CloudTrail, Azure Activity Log) rather than network-layer visibility.
- T1556partialdetects — A.8.20's network-focused monitoring (clause f, j) can surface some downstream signals of T1556 — unusual authentication patterns to network services, connections from systems whose auth process has been tampered with, and modifications to network device authentication — but the core mechanism (LSASS/SAM/PAM tampering on hosts) is host-based and outside the network monitoring scope, so what is caught is a slice of downstream effects rather than the modification itself.
- T1556.001partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which contains the adversary's ability to use the injected skeleton key over the network, though it cannot eradicate the in-memory patch on the domain controller itself.
- T1556.004partialprevents — A.8.20's hardening of network devices (k) and authentication of systems on the network (h) can include practices that make system-image tampering harder (verified boot, signed images, configuration baseline integrity), but the clause's primary focus is network-level controls (filtering, segregation, connection authentication) rather than firmware/OS-image integrity verification, so it constrains rather than forecloses the implant.
- T1557partialresponds — A.8.20 has genuine response mechanisms against an active AiTM — (m) isolating critical subnetworks under attack, (i)/(j) filtering and restricting the rogue device's connection, (n) disabling the abused protocol (e.g. LLMNR, ARP) — which are real containment and blocking actions once the technique is underway; but the clause is a network-security control set, not an incident-response procedure, so it does not cover assessment, evidence collection, stakeholder communication, credential/session-token revocation, or root-cause analysis, leaving the response a slice of the full task set.
- T1557.001partialdetects — A.8.20(f) explicitly mandates logging and monitoring to enable detection of actions relevant to information security, and (j) calls out detecting connections of equipment — both well-positioned to surface anomalous LLMNR/NBT-NS/mDNS responses on the wire. However, the clause sets scope rather than mandating instrumentation depth: an implementation monitoring only perimeter flows or basic connectivity is fully conformant and may miss poisoning traffic on the local segment, so the grade follows the A.8.16 anchor pattern.
- T1557.001partialresponds — A.8.20's network-layer mechanisms (firewalls per (i), subnetwork isolation per (m), device authentication/restriction per (j)) can directly contain an active poisoning attack by blocking the attacker's traffic, isolating the affected segment, or disconnecting the rogue device — but the response is bounded to network-layer actions and does not address credentials already captured, systems already compromised via SMB relay, or the procedural response activities (evidence collection, communication, root cause) that A.5.26 covers.
- T1557.002partialprevents — A.8.20 covers the mechanisms that prevent ARP cache poisoning — hardening of network devices (k) including Dynamic ARP Inspection, authenticating systems on the network (h) via 802.1X, restricting connection of equipment (j), and disabling vulnerable protocols (n) — but ARP is fundamental to IPv4 and cannot be fully disabled, and the clause sets requirements rather than mandating these specific mechanisms, so implementations vary and the prevention is a slice rather than a closure.
- T1557.003partialprevents — A.8.20 requires disabling vulnerable network protocols and restricting/filtering system connections, which can block DHCP spoofing at the network edge, but the control's scope is broad and does not mandate DHCP snooping or specific anti-spoofing mechanisms, leaving implementations that omit those measures fully conformant and vulnerable.
- T1557.004partialprevents — A.8.20 includes mechanisms that partially prevent evil twin attacks — 802.1X authentication (item h) prevents devices from associating with rogue APs lacking proper auth, WIDS/NAC (item j) detects and restricts rogue equipment, and hardening/disabling vulnerable protocols (items k, n) reduces the attack surface — but the attack fundamentally occurs outside the organization's network perimeter on adversary-controlled infrastructure, and A.8.20 does not address client-side certificate validation, PNL behavior, or user decisions to manually connect to rogue APs, so a meaningful slice of the technique remains reachable.
- T1558.001partialdetects — A.8.20's item (f) requires logging and monitoring of network actions, which can surface some Golden Ticket indicators (unusual Kerberos traffic volumes, anomalous authentication patterns to domain controllers), but the primary detection requires Kerberos protocol-level analysis at the domain controller (ticket lifetime anomalies, encryption-type anomalies) that is outside A.8.20's network-level scope.
- T1558.002partialdetects — A.8.20's logging/monitoring provisions (item f) and connection detection (item j) can surface some network-level indicators of silver ticket abuse (unusual service access patterns, traffic anomalies), but the core detection of forged Kerberos TGS tickets requires authentication-layer monitoring — ticket signature validation, encryption-type analysis, SPN checks — which is outside A.8.20's network-security mechanism and belongs to A.8.15/A.8.16.
- T1558.003partialdetects — A.8.20's item (f) logging/monitoring and item (j) connection detection can surface anomalous TGS request volumes or patterns to a domain controller, but Kerberoasting abuses legitimate Kerberos functionality so individual requests are indistinguishable from normal use, and the offline cracking component is invisible to network monitoring entirely.
- T1558.004partialdetects — A.8.20 item (f) requires network logging and monitoring for actions relevant to information security, which can surface the LDAP enumeration queries and AS-REQ/AS-REP patterns that AS-REP Roasting generates, but the clause sets a general network-monitoring scope rather than mandating Kerberos-aware detection, and the offline password-cracking component of the technique is not network-observable at all.
- T1561partialdetects — A.8.20's network logging/monitoring (item f) and network device hardening (item k) can detect worm-like propagation across the network and configuration wipes on network devices, but the core disk-wipe action on hosts is host-level disk I/O that is not visible at the network layer the control governs.
- T1561partialprevents — A.8.20's network controls (segmentation, firewalls, authentication, network device hardening, disabling vulnerable protocols) can prevent the worm-like network propagation variant of T1561 and constrain the network-device CLI wipe variant, but the core technique — direct write access to a disk on a system the adversary has reached — is not stopped by network-layer controls, so what is prevented is a slice of the technique's reach rather than the technique itself.
- T1561partialresponds — A.8.20 item (m) — temporarily isolating critical subnetworks with drawbridges if the network is under attack — is an explicit containment response, and items (i), (j), (n) can also bound worm-like propagation of disk-wipe malware across the network; however, A.8.20 is primarily a preventive control and does not address the disk wipe itself on already-compromised systems, so the response is a slice of the technique's full scope.
- T1561.001partialdetects — A.8.20's network logging and monitoring (item f) and connection detection (item j) can surface the worm-like propagation across networks that T1561.001 describes (lateral movement via Valid Accounts, SMB/Admin Shares), but the core mechanism — overwriting disk contents via direct storage access or RawDisk-style drivers — is a host-level storage operation outside any network-security control's vantage point, so the technique's primary action is not detected.
- T1561.001partialresponds — A.8.20(m) explicitly provides for temporarily isolating critical subnetworks (drawbridges) when the network is under attack, and (i)/(j) restrict and filter connections — these are real containment actions against the worm-like propagation variant of T1561.001. However, the core technique is local disk overwrite, which network controls neither contain on an already-affected host nor eradicate; the wiping mechanism and the wiped content are outside A.8.20's scope.
- T1561.002partialresponds — A.8.20 contains genuine response capability for the worm-like network propagation aspect of T1561.002 — item (m) explicitly calls for temporarily isolating critical subnetworks (drawbridges) when the network is under attack, and items (i) and (j) restrict and filter connections to bound lateral spread — but this is a slice of the control's purpose, which is primarily preventive, and it does not address the core disk-structure impact already realised on infected systems.
- T1561.002partialprevents — A.8.20's network controls (segmentation, firewalls, connection filtering, authentication of systems, disabling vulnerable protocols, isolating critical subnetworks) can limit the worm-like propagation that T1561.002 explicitly leverages via valid accounts, credential dumping and SMB shares, but cannot prevent the disk-structure write itself on any system the adversary has already reached — the technique's core mechanism is a local disk operation, not a network one.
- T1563partialdetects — A.8.20 item f) explicitly requires logging and monitoring to detect actions affecting information security, and items i/j/m/n plus network diagrams and virtualized network separation provide observable signals (e.g. anomalous connections, protocol use, isolation events) that can surface session hijacking in flight; this is limited to a slice because the clause sets scope by organizational requirements rather than mandating universal depth or coverage of all hijack vectors (e.g. in-memory RDP/SSH manipulation with no network artifact).
- T1563partialresponds — A.8.20 includes explicit response-capable items — (m) temporarily isolating critical subnetworks under attack, (i) filtering/restricting connections, (l) segregating admin channels — that can contain a hijacked session and limit lateral movement, but it is primarily a preventive architecture control and does not carry the incident response procedure (A.5.26), assessment (A.5.25), or evidence collection (A.5.28), so it performs a slice of the response tasks rather than the bulk.
- T1563.002partialdetects — A.8.20's logging and monitoring (item f) and network-level controls can surface some indicators of RDP hijacking — unusual session patterns, lateral movement via RDP, connection anomalies — but the core hijack action (tscon.exe execution) is a host-level event not directly visible to network controls, and a hijacked session continues to look like legitimate RDP traffic on the wire.
- T1564.006partialdetects — A.8.20's network monitoring (item f) and connection detection (item j) can see traffic from VMs and flag new device connections, but T1564.006 is specifically designed to evade network-based detection by hiding activity inside the virtual instance where network monitoring cannot see, and by making VM traffic attribution to the host difficult.
- T1565partialdetects — A.8.20's network logging and monitoring (item f) and connection filtering can surface some indicators of data manipulation — unusual database connections, anomalous data flows, or in-transit tampering — but the bulk of T1565 involves manipulating stored data or application-layer data through legitimate access, which network-layer controls do not see.
- T1565partialprevents — A.8.20 explicitly addresses safeguarding the integrity of data passing over networks (item e), and supports this with protocol hardening, disabling vulnerable protocols, network segmentation, and authentication of systems on the network — collectively preventing network-transit manipulation such as MITM-style data tampering. It does not address data manipulation occurring at the application, database, or storage layer, nor manipulation by an actor who has already gained host-level access, so the technique's broader surface remains reachable.
- T1565.002partialdetects — A.8.20(f) explicitly calls for logging and monitoring to detect actions affecting information security, and (e) addresses confidentiality and integrity of data passing over networks, so network-borne manipulation is within the clause's detection scope; however, T1565.002 also operates between system processes where A.8.20 has no vantage point, and even on-network manipulation requires specific integrity-checking depth that the clause does not mandate.
- T1565.002partialprevents — A.8.20 item (e) explicitly addresses safeguarding the integrity of data passing over networks, and items (i), (k), (l), (n) provide firewalls, hardening, segregation and protocol controls that collectively prevent much in-transit manipulation; however, T1565.002 also covers manipulation between system processes on the same host, which network-layer controls cannot reach, leaving a bounded remainder.
- T1565.002partialresponds — A.8.20 contains explicit response actions — item (m) isolating critical subnetworks via drawbridges 'if the network is under attack' and item (n) disabling vulnerable protocols — which can contain an in-flight data manipulation by severing the attacker's path. However, the clause's response is limited to network-level containment; it does not address eradication of the intercepting tool, evidence collection, integrity restoration, or stakeholder communication, so it performs only a slice of the response set.
- T1566partialdetects — A.8.20's network-layer monitoring (item f) and connection controls (items i, j) can detect some phishing-related network activity — outbound connections to known malicious infrastructure after a click, anomalous traffic patterns, or C2 callbacks — but the core of T1566 is the socially-engineered message arriving in a user's inbox, which is an email/application-layer event outside A.8.20's network-security scope.
- T1566.002partialdetects — A.8.20's network monitoring, firewall logging and connection filtering (clauses f, i, j) can detect the post-click network activity — outbound connections to malicious infrastructure, suspicious downloads, and C2 — but the spearphishing email delivery itself and the user's click are outside the network-security scope, so what is caught is a slice of the technique rather than the bulk.
- T1566.003partialdetects — A.8.20's network monitoring (item f) can surface some downstream artifacts when the target clicks the malicious link on a work computer — outbound connections to suspicious destinations, DNS queries, unusual traffic to non-corporate services — but the core technique operates through third-party platforms (social media, personal webmail) explicitly chosen to bypass enterprise network controls, so the social engineering and payload delivery themselves are outside the network's primary visibility.
- T1567partialdetects — A.8.20 item (f) requires logging and monitoring of network activity to enable detection of actions relevant to information security, and T1567 exfiltration over web services does generate observable network traffic (volumes, timing, destinations) on the organization's estate; however, the technique specifically uses legitimate services with SSL/TLS encryption, so content-based detection at the network level is largely unavailable and what is caught is a behavioral slice rather than the bulk of the technique.
- T1567partialprevents — A.8.20's network segmentation, egress filtering (i), and restriction of system connections can prevent some exfiltration over web services by limiting which systems reach which external destinations, but the technique is specifically designed to use legitimate, already-permitted services over encrypted channels, so the control cannot reach the bulk of the technique.
- T1567partialresponds — A.8.20's drawbridge mechanism (m) and filtering/restriction capabilities (i, j) provide containment actions that can be triggered once exfiltration is underway — isolating the affected subnetwork or blocking the connection stops further data leaving. However, the clause is primarily preventive (firewalls, protocol hardening, authentication) and the response capability is limited to network-level isolation; it does not include eradication of the foothold or the broader incident-response activities (assessment, evidence collection, closure) that A.5.26 carries.
- T1567.001partialdetects — A.8.20(f) explicitly mandates logging and monitoring of network actions, and the exfiltration traffic does traverse the network where flow data, destination, volume and timing are observable; however, T1567.001 is specifically designed to evade network-level detection by riding HTTPS to a legitimate, often pre-existing code-repository service, so content is opaque and the signal is confounded with normal API use.
- T1567.001partialprevents — A.8.20's network filtering, egress controls and subnet isolation can block some exfiltration paths to code repositories, but T1567.001 uses standard HTTPS to legitimate services (GitHub, GitLab) that organizations typically allow, so the technique is not prevented where the destination is permitted and the traffic is encrypted — only a slice of paths is closed.
- T1567.001partialresponds — A.8.20 includes explicit response-capable elements — drawbridge isolation of subnetworks 'if the network is under attack' (item m) and firewall filtering that can be applied to cut off an exfiltration channel once underway — but it is a network-security control, not an incident response procedure, so it covers only network-level containment and not assessment, evidence collection, eradication of the foothold, or closure; for T1567.001 specifically the HTTPS-to-code-repository channel is hard to distinguish from legitimate traffic, narrowing what network-level response can actually do.
- T1567.002partialdetects — A.8.20's network monitoring, logging and firewall controls (items f, i, j) can detect the outbound connection to a cloud storage service and surface traffic volume/timing anomalies, but the technique is designed to ride legitimate encrypted HTTPS to known cloud services, so content is opaque without TLS interception and the connection itself blends with normal cloud use — a slice is caught, not the bulk.
- T1567.002partialprevents — A.8.20's firewall/filtering (i), segregation (l, m), and connection-restriction mechanisms can block egress to cloud storage services, but the technique is explicitly designed to blend with traffic that hosts are 'already communicating with' — legitimate cloud service use is widespread, so blocking is selective, and adversaries can use personal accounts, alternative services, or encrypted channels that defeat simple port/filter rules.
- T1567.002partialresponds — A.8.20 provides the capability to temporarily isolate critical subnetworks or filter connections when under attack, which contains the exfiltration in progress but does not eradicate the adversary's foothold or staged data.
- T1567.003partialdetects — A.8.20's logging and monitoring (item f) and connection filtering (items i, j) can surface outbound traffic to text storage sites — destination, volume, timing — but the clause does not mandate content inspection or DLP, so distinguishing exfiltration from legitimate use of these services is not guaranteed by the control's mechanism.
- T1567.003partialprevents — A.8.20's firewall and filtering guidance (i, j) and protocol restrictions (n) can block egress to known text storage sites, but the clause is oriented toward network architecture and ingress protection rather than mandating egress filtering, and adversaries can rotate across many text storage services over HTTPS that firewalls commonly allow.
- T1567.003partialresponds — A.8.20 has response capability through (m) temporarily isolating critical subnetworks and (i,j) restricting/filtering connections, which can contain an in-progress exfiltration to a text storage site, but the clause is primarily preventive and detective; eradication of the actor's foothold is outside its scope, and the response elements are a minority of the clause's overall function.
- T1567.004partialdetects — A.8.20 item (f) explicitly mandates network logging and monitoring for detection, and the network is a genuine vantage point for HTTPS traffic to webhook endpoints; however, the technique is specifically designed to blend with normal SaaS traffic over encrypted channels, so network-level detection relies on traffic-pattern anomalies (volume, frequency, destination reputation) rather than content inspection, and A.8.20 sets the monitoring framework without mandating the depth (DLP, TLS inspection) needed to reliably distinguish webhook exfiltration from legitimate use.
- T1567.004partialprevents — A.8.20's network controls (firewalls, egress filtering, network segmentation, segregation of administration channels) can partially prevent webhook exfiltration by blocking known webhook endpoints or restricting which systems can reach external SaaS services, but the technique is specifically designed to use legitimate HTTPS to commonly used services (Discord, Slack) that firewalls must allow for business operations, so the bulk of the technique passes through these controls.
- T1567.004partialresponds — A.8.20 includes explicit reactive network actions — (m) temporarily isolating critical subnetworks under attack and (i) restricting/filtering connections — that can contain an exfiltration already underway by severing the affected system's network access or blocking the webhook traffic. However, the clause is primarily a preventive control (firewalls, hardening, protocol disabling, segregation) and its response toolkit is narrow: it acts at the network layer only, has no procedure for evidence collection or eradication of the foothold, and webhook exfiltration over HTTPS to legitimate SaaS endpoints is difficult to distinguish from normal traffic, so the filtering that does engage is a slice rather than a bounded remainder of a working mechanism.
- T1568partialprevents — A.8.20's network filtering, firewall controls, and restrictions on outbound connections can block some dynamic resolution attempts (e.g., DNS sinkholing, blocking known malicious resolvers, restricting outbound DNS), but the technique is specifically designed to evade network-based controls by using legitimate DNS infrastructure, algorithmically generated domains, and public resolvers, so the control prevents a slice rather than the bulk.
- T1568partialresponds — A.8.20 can perform containment actions against T1568 by blocking dynamically resolved C2 traffic at the firewall and isolating affected subnets (per items i, j, m), but it does not eradicate the malware or the dynamic resolution mechanism on the endpoint, leaving the response incomplete.
- T1568.001partialdetects — A.8.20(f) requires logging and monitoring of network actions to enable detection, and DNS queries/responses from internal hosts to fast-flux domains are observable in that traffic; however, the clause sets general monitoring scope rather than mandating DNS-specific analysis or the pattern detection (many A records, short TTLs, rapid IP churn) that fast flux requires, so a conformant implementation focused on connection-level monitoring could miss it.
- T1568.002partialdetects — A.8.20's item (f) mandates network logging and monitoring, and DGAs produce observable DNS traffic patterns (high query volumes, NXDOMAIN responses, algorithmically generated names) that traverse the network — so the technique is in scope of the clause's monitoring requirements, unlike T1583.001 which happens entirely off-estate. However, A.8.20 sets a general monitoring framework rather than mandating specific DGA detection capabilities; identifying algorithmically generated domains requires analytical depth (statistical analysis, ML, threat intel feeds) that an implementation conformant with A.8.20 may or may not possess, and encrypted DNS (DoH/DoT) can evade the network vantage point.
- T1568.002partialprevents — A.8.20's firewalls, egress filtering, DNS controls and network segmentation can block or sinkhole some DGA-issued destinations and constrain where C2 traffic can reach, but the technique is explicitly designed to defeat blocklists by generating thousands of candidate domains, and the algorithm itself runs on the endpoint where network controls have no reach — so a slice of DGA C2 is prevented, not the technique.
- T1568.002partialresponds — A.8.20 item (m) explicitly provides containment through temporary isolation of critical subnetworks (drawbridges) when the network is under attack, which is a genuine response action against an active DGA C2 channel; however, the clause is primarily preventive/detective architecture and does not perform eradication, evidence collection, or recovery, so the response is limited to network-level containment of the channel rather than the full response lifecycle.
- T1569partialdetects — A.8.20's monitoring scope is network-level (traffic, connections, network devices per clauses f, i, j), so it can surface remote service abuse variants of T1569 (e.g. PsExec/SMB, RPC/SCM remote service creation) through anomalous administrative traffic, but the local service/daemon abuse variants — which constitute the bulk of the technique — execute entirely on the endpoint and produce no network signal A.8.20 is positioned to see.
- T1569.002partialprevents — A.8.20's network controls (firewalls, segmentation, filtering, disabling vulnerable protocols) can block the remote execution path of T1569.002 — PsExec and sc.exe rely on SMB/RPC traffic that network restrictions can refuse — but the technique also runs locally via services.exe on a compromised host, which no network control reaches.
- T1570partialprevents — Network segmentation (m), firewalls (i), disabling vulnerable protocols (n), and system authentication (h) can prevent some lateral tool transfer between segments or via specific channels, but the technique uses many legitimate protocols (SMB, RDP, scp, rsync, curl, sftp, ftp, cloud sync) that are often required for normal operations and cannot be removed without breaking the business, so the technique still runs within permitted channels.
- T1571partialprevents — A.8.20 mandates firewalls, filtering, and disabling vulnerable protocols (items i, n, k), which can block non-standard port usage, but T1571 is specifically designed to bypass port-based filtering, and the clause sets requirements rather than mandating deep packet inspection or application-layer filtering needed to fully defeat it.
- T1571partialresponds — A.8.20 includes explicit response mechanisms — drawbridges for isolating subnetworks under attack (m), restricting connections (j), and disabling vulnerable protocols (n) — that can act against non-standard port usage once observed, but the clause is primarily a preventive network-security control and the response capability is a slice of its function rather than its core.
- T1572partialdetects — A.8.20(f) and (j) explicitly require logging, monitoring, and detecting actions affecting information security and connections to the network, which provides the framework for surfacing tunneling; however, T1572 is specifically designed to evade detection by encapsulating traffic within legitimate-looking protocols (SSH, DoH) and providing encryption, so what monitoring catches is a slice — anomalous volumes, unexpected destinations, protocol-pattern anomalies — while well-formed encrypted tunnels that mimic legitimate traffic remain invisible.
- T1572partialprevents — A.8.20's firewalls, protocol filtering (i), disabling vulnerable protocols (n), and network segmentation can block some tunneling vectors (e.g., unnecessary outbound SSH), but the technique is explicitly designed to evade network filtering by encapsulating traffic within permitted protocols like HTTPS and DNS, which cannot be blocked without breaking legitimate services.
- T1572partialresponds — A.8.20(m) explicitly provides a response action — temporarily isolating critical subnetworks (drawbridges) when the network is under attack — which is genuine containment against tunneling once detected; however, the clause is primarily preventive/detective (firewalls, filtering, hardening, logging) and does not itself perform eradication of the tunneling capability or full incident handling, so only a slice of the response verb is delivered.
- T1573partialdetects — A.8.20's logging and monitoring (item f) and connection detection (item j) can observe encrypted C2 traffic as it traverses the network — volumes, timing, endpoints, and connection patterns remain visible — but the technique's defining feature (adversary-controlled encryption of payload) defeats content inspection, so what is detected is metadata and behavioural anomalies rather than the C2 communications themselves.
- T1573partialprevents — A.8.20 requires network segmentation, traffic filtering, and disabling vulnerable protocols, which can block or restrict the adversary's ability to establish an encrypted channel, but it cannot prevent the use of encryption over allowed protocols or channels that remain open for business needs.
- T1573partialresponds — The control includes temporarily isolating subnetworks during an attack, which contains the encrypted command and control channel by severing the connection, but does not eradicate the malware generating it.
- T1573.001partialresponds — Isolating subnetworks or filtering connections contains the encrypted command and control traffic, but does not eradicate the malware performing the encryption.
- T1573.002partialdetects — A.8.20's logging and monitoring (item f) can record network traffic patterns that may indicate C2 activity using this technique — beaconing intervals, unusual destinations, traffic volumes — but the asymmetric encryption itself is designed to be opaque to network inspection, so what is detected is the surrounding behavior rather than the cryptographic method, and the scope of monitoring is set by implementation rather than mandated in depth.
- T1573.002partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which contains the encrypted command and control traffic but does not eradicate the adversary's foothold or key material.
- T1580partialdetects — A.8.20 item (f) requires logging and monitoring of network actions, and cloud API calls (DescribeInstances, ListBuckets, etc.) traverse the network and are visible in that telemetry; however, the calls use valid authentication and resemble legitimate cloud management traffic, so network monitoring alone cannot reliably distinguish malicious discovery from normal operations without additional context such as UEBA or credential-compromise indicators.
- T1590partialdetects — A.8.20's monitoring and logging (item f) detects the active-scanning component of T1590 that touches the organization's network, but the technique also includes passive methods (WHOIS, DNS Dumpster, search open technical databases) that occur entirely off the estate and produce no observable event for the control to surface.
- T1590partialprevents — A.8.20's firewalls, filtering, hardening and protocol disabling can limit active scanning and reduce information disclosure from the network itself, but T1590's vectors include WHOIS, DNS, public databases and phishing for information that occur outside the organization's network and are not stopped by these controls.
- T1590.004partialdetects — A.8.20 explicitly requires logging/monitoring (item f, cross-referencing 8.16) of actions affecting network security, which can surface topology-gathering activity such as active scanning; this is limited to a slice because the control's scope is set by organizational requirements rather than mandating detection of all topology reconnaissance vectors (e.g., passive public data exposure).
- T1590.004partialprevents — A.8.20 requires network segregation, hardening, protocol disabling, authentication of systems, connection filtering (firewalls), logging/monitoring, and temporary isolation under attack, all of which constrain an adversary's ability to gather accurate topology via scanning or exposed data; it does not stop all collection vectors such as phishing for information or public data sets.
- T1590.006partialdetects — A.8.20's logging and monitoring (item f) can detect active scanning that probes the estate for security appliances, but the technique is also performed via phishing for information, open-source research, and other off-estate methods that produce no event A.8.20 can observe.
- T1592partialdetects — A.8.20's logging and monitoring (item f) and connection detection (item j) can observe active scanning and queries to victim-owned infrastructure that touch the estate, but T1592 also encompasses passive methods—social media research, general web searching, open technical database queries—that occur entirely off the organization's network and have no vantage point for A.8.20 to detect.
- T1592.001partialprevents — A.8.20's firewalls, filtering, network segregation and protocol hardening can prevent some active scanning that traverses the network to harvest hardware details, but T1592.001 explicitly enumerates non-network vectors — OSINT (job postings, resumes, purchase invoices, network maps), phishing for information, and compromised-site collection — that no network control reaches, so only a slice of the technique's surface is closed.
- T1592.002partialprevents — A.8.20's network controls (firewalls, filtering, hardening, disabling vulnerable protocols) can prevent some software fingerprinting via network scanning (banners, ports, user agents), but T1592.002 also draws on public sources (job postings, network maps, metadata in hosted files) and phishing for information, which network controls do not reach.
- T1592.004partialprevents — A.8.20's firewalls, filtering, hardening and protocol disabling can prevent or limit active-scanning vectors of T1592.004 (port scans, banner grabs, user-agent harvesting) and reduce what client configuration data leaks through network services, but the technique has multiple vectors A.8.20 does not reach — phishing for information, compromised sites collecting host data, and online data sets (job postings, network maps, resumes, purchase invoices) — so only a slice of the technique's information-gathering surface is prevented.
- T1595.001partialdetects — A.8.20 requires logging and monitoring of network actions, which can record the inbound ICMP and TCP probes of an IP block scan, but the control's scope is the organization's own network and it cannot observe scans that target unallocated addresses or occur entirely outside its perimeter.
- T1595.001partialprevents — A.8.20's filtering, hardening, and protocol-disabling controls prevent much of the detailed information a scan would harvest (services, banners, versions), but the scan itself still runs — packets arrive at the perimeter, the IP block's existence and basic responsiveness are still observable, and the adversary can still initiate the probe — so the technique is constrained, not stopped.
- T1595.001partialresponds — A.8.20 provides the network filtering and isolation mechanisms used to dynamically block scanning IPs once detected, containing the probe on the estate while the adversary's broader reconnaissance campaign continues off-estate.
- T1595.002partialprevents — A.8.20 prevents a slice of vulnerability scanning through firewalls (i), network segmentation (m), and disabling vulnerable protocols (n), but the technique still runs against any public-facing service the organization must expose, and the adversary's probes reach the perimeter regardless of internal controls.
- T1595.002partialresponds — A.8.20 has direct technical response mechanisms — firewalls (i), connection filtering (j), and subnetwork isolation via drawbridges (m) — that can block or contain the scanning source once it is underway, and these are genuine containment actions against an active scan. The remainder is the same one named on the A.5.26 vs T1595.002 anchor: the response reaches the contact, not the campaign; the adversary can resume from other sources and the information already harvested cannot be eradicated.
- T1595.003partialdetects — A.8.20 item (f) mandates logging and monitoring of actions relevant to information security, and wordlist scanning against the organization's own web servers and cloud endpoints generates observable traffic patterns (many requests to varied paths) that network devices, firewalls and web server logs can surface; however, A.8.20 is a broad network-security control whose primary purpose is protection rather than detection, and the extent of detection depends on implementation specifics such as whether IDS/WAF rules are tuned for scanning patterns.
- T1595.003partialprevents — A.8.20's firewalls, segmentation, authentication and protocol restrictions prevent wordlist scanning against internal and non-public systems, but public-facing services (web servers, cloud storage buckets) must remain reachable to serve legitimate traffic, and scanning requests are indistinguishable from normal ones — so the technique cannot be prevented against the public-facing surface that is its primary target.
- T1599partialdetects — A.8.20's guidance item (f) explicitly requires logging and monitoring of network devices to enable recording and detection of actions affecting information security, which can surface indicators of T1599 such as configuration changes on boundary devices and anomalous traffic patterns crossing trust boundaries; however, the technique specifically compromises the very routers/firewalls the control protects, and once subverted the adversary can evade device-level logging, so detection is not guaranteed and depends on external monitoring depth.
- T1599partialprevents — A.8.20 mandates network hardening, authentication, and traffic filtering that reduce the attack surface for compromising a boundary device, but a determined adversary with sufficient resources or a zero-day can still breach it, so prevention is a reduced likelihood rather than a closure.
- T1599partialresponds — A.8.20 item (m) provides a genuine containment response — temporarily isolating critical subnetworks via drawbridges when the network is under attack — which applies directly to T1599's bridging of compromised boundary devices; however, the clause does not cover eradication of the adversary from the compromised device, evidence collection, or restoration of device configuration, so the response is a slice (network-level isolation) rather than the bulk of what responding to this technique requires.
- T1599.001partialprevents — A.8.20 requires network segregation, device hardening, and disabling of vulnerable protocols, which collectively reduce the attack surface for gaining control of a boundary device to modify NAT, but the control does not eliminate the possibility of compromise through zero-days or misconfiguration.
- T1599.001partialresponds — A.8.20 has genuine response actions against an in-flight NAT traversal — (m) temporarily isolating critical subnetworks via drawbridges and (j) restricting connections to the network are real containment moves once the technique is underway — but these are a slice of the full response function; eradication of the adversary's foothold on the compromised boundary device, evidence collection, and incident coordination belong to A.5.26 / A.5.28, not this clause.
- T1600partialdetects — A.8.20(f) requires logging and monitoring of actions relevant to information security across the network, and (j) explicitly detects connection of equipment and devices — so a vantage point on network devices exists and weakened encryption can surface as anomalous traffic patterns or configuration drift. But the technique compromises the device's own encryption capability (system image modification, key-space reduction, crypto-hardware disablement), which is internal to the device and requires specific integrity or configuration monitoring that A.8.20 does not mandate; scope is set by implementation rather than by the clause, so what is caught is a slice.
- T1600partialresponds — A.8.20(m) provides for temporarily isolating critical subnetworks (drawbridges) if the network is under attack, which is a containment response once encryption on a network device has been weakened; however, the clause does not cover eradication of the compromise on the device (re-imaging, restoring encryption capability) or the broader incident-response activities, so only a slice of the response is performed.
- T1600.001partialdetects — A.8.20(f) requires logging and monitoring of actions affecting information security, and (k)/(j) address network device hardening and connection monitoring, providing mechanisms that could surface a key-size reduction on a compromised device via configuration-change logs or traffic analysis — but the clause sets scope rather than mandating specific detection of cipher-strength changes, and a device already compromised by the adversary may evade local logging, so detection depends on implementation depth.
- T1600.001partialprevents — A.8.20's hardening of network devices (k), disabling vulnerable protocols (n), and configuration management aspects can constrain the ability to weaken encryption on network devices, but the technique presupposes the device is already compromised via Modify System Image, and once an adversary has administrative access, configuration changes including key-size reduction are typically achievable; the control reduces the surface but does not eliminate the technique.
- T1600.002partialprevents — A.8.20's hardening of network devices (k), authentication (h), and restriction of connections (i) can make the initial compromise of a network device harder, which is a prerequisite for disabling its hardware crypto; however, the clause does not specifically mandate preventing configuration changes to crypto settings on already-compromised devices, and the technique runs once the adversary has device control.
- T1601partialdetects — A.8.20's monitoring and logging requirements (item f) and network device hardening (item k) provide a framework that could surface some indicators of OS image modification on network devices (anomalous behavior, configuration changes, administrative access logs), but the clause does not mandate firmware/OS integrity verification specifically, and T1601 is designed to be stealthy by modifying a single monolithic file on the device itself.
- T1601.001partialdetects — A.8.20(f) calls for logging and monitoring of actions affecting information security, and (j) explicitly mentions detecting connections of equipment — so network-level monitoring can surface some indicators of T1601.001 (e.g., anomalous TFTP/FTP/SCP transfers to network devices carrying image files). However, the clause sets scope rather than mandating device-image integrity checks, and in-memory patching via debug commands, boot loader manipulation, and console-based image writes are not visible to network monitoring at all, leaving a substantial remainder.
- T1601.001partialprevents — A.8.20 requires hardening of network devices, disabling vulnerable protocols, and restricting connections, which collectively reduce the attack surface for patching the system image, but the control does not eliminate administrative access or bootloader manipulation entirely, leaving a bounded remainder.
- T1601.002partialdetects — A.8.20(f) requires logging and monitoring of actions relevant to information security, and (j) explicitly mentions detecting connections of equipment — both of which can surface an OS image downgrade on a network device if device-level logs are collected and integrity of the running image is monitored. However, the clause sets a monitoring framework rather than mandating file-integrity or image-version checks specifically, so an implementation focused on traffic-level monitoring is fully conformant and would miss the in-device file replacement and reboot that constitute the technique.
- T1601.002partialprevents — A.8.20's hardening of network devices (k), restriction of connections (j), and disabling of vulnerable protocols (n) are directly relevant and, where faithfully implemented with image signing/integrity verification and secure boot, would prevent an unsigned or unauthorized downgrade; however, the clause sets broad hardening requirements rather than mandating the specific mechanism (signed images, verified boot) that defeats T1601.002, so a fully conformant implementation that hardens devices without enforcing image integrity leaves the technique reachable.
- T1602partialdetects — A.8.20's clauses (f) logging/monitoring and (j) detecting connections of equipment to the network provide the framework to surface anomalous queries to configuration repositories (e.g. SNMP abuse), but the clause sets requirements rather than mandating instrumentation depth, so detection depends on implementation scope and some vectors (e.g. authorized-looking access to a misconfigured repository) may not be caught.
- T1602partialprevents — A.8.20 requires network device hardening, disabling vulnerable protocols, and restricting system connections, which can prevent exposure of configuration repositories, but the control's scope is broad and does not mandate specific protections for every repository protocol or access method.
- T1609partialdetects — A.8.20's network monitoring and logging (item f), authentication of systems on the network (item h), and connection filtering (items i, j) can detect connections to container administration services (Docker daemon, Kubernetes API server, kubelet) and unauthorized access attempts, but cannot see the actual command execution within containers, and authenticated sessions to admin endpoints look legitimate at the network layer.
- T1609partialprevents — A.8.20's network restrictions, filtering, authentication of systems on the network, and segregation of network administration channels can prevent unauthorized remote access to container administration services (Docker daemon, Kubernetes API server, kubelet), but the technique can still execute through local sockets, already-authenticated sessions, or from within a compromised container where network controls do not reach.
- T1610partialdetects — A.8.20's network monitoring (item f) and connection detection (item j) can observe traffic to container orchestration APIs and new container network connections, providing signals of deployment activity, but the actual deployment action occurs at the orchestration/application layer where network-level monitoring has limited semantic visibility into API call content and container configuration.
- T1612partialdetects — A.8.20's network monitoring (item f) and connection filtering (items i, j) can detect a Docker API build request when the daemon is exposed over the network, but T1612 is specifically designed to build locally on the host to bypass such defenses, and a local socket build leaves no network artifact for A.8.20 to observe.
- T1612partialprevents — A.8.20's firewall rules, network segmentation, and authentication of network connections (items h, i, j, l) can prevent unauthorized remote systems from reaching the Docker API to issue a build request, but the technique can still execute from a foothold already inside the network (e.g., a compromised container with access to the Docker socket/API), and A.8.20 does not specifically address container runtime or Docker API hardening.
- T1613partialdetects — A.8.20's logging/monitoring (item f) and connection detection (item j) can surface API calls to Kubernetes/Docker control planes and dashboard access from unexpected sources, but container discovery via local sockets or in-container CLI tools (kubectl, docker) does not generate network traffic the controls can see, leaving a meaningful slice undetected.
- T1613partialprevents — A.8.20's network access controls (firewalls, authentication, segmentation per items i, j, m) can block external or unauthorized access to container APIs and dashboards, preventing some instances of T1613, but the technique is typically executed from within a compromised container where perimeter network controls do not apply and the API calls traverse the cluster's internal network.
- T1614partialprevents — A.8.20's network controls (firewalls, egress filtering, network segregation) can block the IP-geolocation lookup vector of T1614 by restricting outbound traffic to geolocation services, but the host-based discovery vectors — time zone, keyboard layout, language settings, Windows API calls like GetLocaleInfoW, and cloud instance metadata queries — execute entirely on the endpoint and are not reachable by network-layer controls.
- T1615partialdetects — A.8.20's network logging and monitoring (item f) can surface anomalous LDAP queries to domain controllers that carry the GPO discovery, but the technique's core execution — `gpresult`, `Get-DomainGPO` and similar PowerShell running on a compromised host — is host-level activity outside A.8.20's network scope, so what is caught is the transport rather than the act.
- T1619partialdetects — A.8.20's network monitoring scope (item f) can observe API calls traversing the organization's network to cloud storage endpoints, surfacing enumeration patterns by volume, source or destination, but the natural and primary detection for T1619 is cloud-side audit logging (CloudTrail, Storage Logs) rather than network monitoring, and encrypted API payloads limit semantic visibility into what is being enumerated.
- T1621partialdetects — A.8.20's network monitoring (items f, j) can observe the repeated authentication requests and unusual connection patterns that accompany MFA fatigue, but the core of T1621 — the push notification generation at the IdP and the user's acceptance on their device — happens outside the network monitoring scope, so only the precursor traffic is detectable.
- T1654partialdetects — A.8.20's network monitoring scope (item f) and connection filtering (items i, j) can surface network-level signals of T1654 — bulk log exfiltration to adversary-controlled infrastructure, unusual access patterns to centralized SIEM/logging systems — but the core technique is host-level log access via wevtutil.exe, PowerShell, or local file reads, which falls outside A.8.20's network-centric monitoring scope.
- T1659partialdetects — A.8.20's logging and monitoring (item f) can surface downstream anomalies and effects of injected content arriving at the network, but the core injection mechanism occurs upstream (ISP, lawful interception, man-on-the-side) where the organization's network controls have no vantage point — the traffic appears to come from legitimate sources, so distinguishing injected from legitimate content is fundamentally limited.
- T1659partialprevents — A.8.20's filtering (i), authentication (h), protocol hardening (n), and integrity controls for data over public networks (e) prevent some content injection vectors, but the technique explicitly operates at upstream channels (e.g., ISP-level 'lawful interception') and in-transit manipulation that the organization's network controls cannot reach, leaving a substantial remainder.
- T1659partialresponds — A.8.20 provides genuine response mechanisms — clause (m) isolating subnetworks under attack and clause (i) firewall filtering can contain ongoing content injection — but the injection point is typically upstream (ISP or network path) outside the organization's control, so the response reaches the affected subnetwork and filters traffic at the perimeter while leaving the upstream channel untouched, and the malicious content already delivered to victim systems is not undone by network-level response.
- T1665partialdetects — A.8.20's logging and monitoring (item f) and network traffic controls can surface some artifacts of T1665 — anomalous proxy/VPN patterns, IP spoofing attempts, and filtering of security-tool traffic — but the core of the technique (infrastructure setup, domain masking, use of trusted hosting) happens on adversary/registrar infrastructure or is designed to blend with normal traffic, so most of the hiding is outside the control's vantage point.
- T1669partialdetects — A.8.20 requires logging and monitoring to enable recording and detection of actions affecting information security, which can surface anomalous wireless connections, but the control's scope is the organization's own network infrastructure and does not mandate the wireless-specific monitoring (e.g., rogue access point detection, wireless intrusion prevention) needed to fully detect unauthorized Wi-Fi connections.
- T1669partialresponds — A.8.20 includes explicit response mechanisms for an attack already underway — item (m) temporarily isolating critical subnetworks under attack, item (i) restricting and filtering connections, item (j) detecting and restricting unauthorized device connections — which can contain an adversary who has gained Wi-Fi access. However, the clause does not cover eradication of the foothold, investigation, or recovery, so the response is limited to network-level containment.
- T1673partialdetects — A.8.20's item (f) logging/monitoring covers network traffic, so VM discovery via vCenter GUI or remote management sessions over the network is observable, but the primary mechanism — local CLI enumeration on the hypervisor (esxcli, vim-cmd) — does not traverse the network and falls outside A.8.20's network-scoped monitoring.
- T1675partialdetects — A.8.20's logging and monitoring scope (item f) and network traffic filtering (items i, j) can surface anomalous traffic to/from ESXi management interfaces and guest VMs, but T1675 is fundamentally abuse of legitimate vSphere administration APIs at the management plane, which A.8.20 does not directly instrument; detection is a slice of the technique's observable surface.
- T1675partialprevents — A.8.20's network controls (firewalls, segmentation of administration channels per item l, authentication per item h, filtering per item i) can prevent unauthorized network access to ESXi management interfaces, but the technique operates via legitimate VMware Tools and vSphere APIs between the hypervisor and guest VMs — once ESXi is reached, traditional network controls do not stop the abuse of administration services, so prevention is partial and limited to the access path.
- T1675partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which provides network-level containment by cutting off the adversary's remote access to the ESXi management interface, though it does not eradicate the foothold on the host or guest VMs.
- T1685partialdetects — A.8.20's network monitoring (item f), NAC (item j), and device hardening (item k) can detect the network-relevant slice of T1685—syslog/SIEM forwarding disruption, network device tampering, and unauthorized equipment connections—but the bulk of the technique (host-based EDR/antivirus tampering, registry modifications, process killing, ETW/Sysmon interference) falls outside A.8.20's network scope, so only a portion is observable.
- T1685partialprevents — A.8.20's network controls (firewalls, segmentation, network device hardening, disabling vulnerable protocols, restricting connections) can prevent some network-mediated aspects of T1685 — particularly tampering with network-resident security tools like IDS, SIEM ingestion paths, and network device defenses — but the technique's core mechanisms (killing local processes, modifying Registry keys, disabling ETW, deleting host config files, tampering with EDR/AV/logging agents) operate at the host level and are outside A.8.20's scope.
- T1686partialprevents — A.8.20 includes preventive items (i firewalls, k hardening, l segregating admin channels, j restricting connection of devices) that constrain who can reach and modify firewall configurations, but the technique specifically targets the firewall mechanism itself and runs once an adversary has sufficient privileges; A.8.20's general network-security controls reduce the likelihood without closing the path, and the clause does not mandate tamper-protection of firewall rule stores.
- T1686partialresponds — A.8.20(m) explicitly provides a containment response — temporarily isolating critical subnetworks via drawbridges 'if the network is under attack' — which bounds an in-progress firewall tampering by severing the affected segment, but the clause is primarily preventive (firewalls, filtering, hardening, segregation) and does not perform eradication, investigation, or restoration of the modified firewall state.
- T1686.001partialdetects — A.8.20's logging and monitoring provisions (f) and firewall management (i) provide mechanisms that could surface cloud firewall modifications through configuration change monitoring and resulting traffic anomalies, but the technique operates via cloud control plane APIs using valid credentials, and A.8.20's scope is set by implementation rather than mandating cloud audit log integration, so detection is a slice rather than comprehensive.
- T1686.001partialprevents — A.8.20 establishes the network firewall controls (item i) and hardens network devices (item k) that the technique targets, but the technique specifically defeats those controls by modifying firewall rules via the cloud management plane using compromised credentials/permissions — a vector A.8.20 does not address, since protecting firewall configuration from authorized modification belongs to identity and access management controls (A.5.16, A.8.2, A.8.5), not to network-security establishment.
- T1686.001partialresponds — A.8.20 item (m) — temporarily isolating critical subnetworks with drawbridges when the network is under attack — is a genuine containment action that bounds the spread of a modified cloud firewall, but the clause is primarily preventive and its response capability is limited to network isolation rather than directly restoring firewall rules or eradicating the adversary's foothold in the cloud control plane.
- T1686.002partialdetects — A.8.20(f) and (j) require logging and monitoring of actions affecting information security, which would surface firewall rule changes on network devices if logs are forwarded off-device; however, the technique specifically targets the network devices themselves, where an adversary with management access can tamper with or disable local logging, so detection is a slice rather than a bounded remainder of a working mechanism.
- T1686.002partialprevents — A.8.20 requires hardening, authentication, and restriction of network devices, which reduces the attack surface for compromising management interfaces, but a determined adversary with valid credentials or an unpatched exploit can still reach the device and modify firewall rules.
- T1686.003partialdetects — A.8.20's logging and monitoring provisions (f) and connection-detection (j) can surface some effects of host firewall modification through anomalous network traffic patterns, but the control's primary scope is network-level security and the Windows host firewall is a host-based control whose rule changes occur on the endpoint, outside A.8.20's direct monitoring vantage point.
- T1687partialdetects — A.8.20's items (f) logging/monitoring and (j) detecting connection of equipment enable observation of network-layer manifestations of T1687 — exploitation of firewalls, routers, or network devices, and anomalous traffic patterns indicating defensive impairment — but the technique's exploitation of endpoint security tools (EDR, AV) and cloud/SaaS defensive infrastructure occurs outside A.8.20's network vantage point, leaving a substantial unobserved remainder.
- T1687partialprevents — A.8.20's hardening of network devices (k), disabling of vulnerable protocols (n), segregation of administration channels (l), and connection filtering (i) reduce the attack surface for exploiting network-layer defensive components like firewalls and routers, but T1687 also targets EDR, antivirus, and other endpoint/cloud defensive tools outside A.8.20's scope, and hardening lowers rather than eliminates exploitation risk.
- T1687partialresponds — The control includes temporarily isolating critical subnetworks when under attack, which provides network-level containment of the exploitation, but it does not eradicate the threat or address the exploitation of endpoint or SaaS defensive tools.
- T1689partialdetects — A.8.20(f) and (j) explicitly call for logging, monitoring and detection of network-relevant actions, which catches the network-protocol portion of T1689 (e.g. HTTPS-to-HTTP stripping, protocol-version downgrades visible on the wire), but the technique also covers application- and system-level downgrades such as reverting PowerShell to a version without Script Block Logging or downgrading the boot manager to bypass Secure Boot, which sit outside the network-security control's scope.
- T1689partialprevents — A.8.20's items (e) safeguarding confidentiality/integrity over networks and (n) disabling vulnerable network protocols directly address the network-protocol portion of downgrade attacks (e.g., HTTPS→HTTP, TLS downgrade), but the technique also covers application- and system-level downgrades such as PowerShell version rollback and boot manager downgrade, which fall outside A.8.20's network scope; even within the network portion, backward-compatibility negotiation can still permit a forced fallback to a less secure mode.
- T1700partialresponds — A.8.20 includes temporarily isolating critical subnetworks when under attack, which provides network-level containment of the exfiltration traffic, though it does not eradicate the malicious prompt or the AI agent's compromised state.
Prevented OWASP Web Top 10 (2025) risks (9)
OWASP Web Top 10 (2025) risk categories this ISO control helps prevent or mitigate — our analysis (mostly llm_verified).
Direction: ← other covers this;
→ this covers other (F/M/P = full / mostly /
partial). gov = governs / implements (a mandate, not coverage).
Why these map — rationale
- A02mostlyblocks — A.8.20 explicitly requires hardening network devices, restricting/filtering connections (e.g. firewalls), segregating admin channels, disabling vulnerable protocols, and applying controls to virtualized/SDN networks, which directly closes most misconfiguration attack surfaces named by A02:2025; the bounded remainder is non-network application/framework defaults and incomplete hardening of endpoints that sit outside the network layer.
- A01partialblocks — A.8.20's network-level mechanisms (firewalls, connection filtering/restriction, segregation, protocol disabling, virtualized isolation) squarely prevent several A01 members that cross network boundaries (e.g. SSRF, path traversal over networks, certain CSRF vectors), but leave the bulk of the category (IDOR, missing function-level checks, authorization logic flaws inside the app) untouched.
- A05partialblocks — A.8.20's network-level measures (firewalls, protocol disabling, segmentation, isolation under attack, virtualized separation) prevent several injection families from reaching their interpreters (e.g. OS command, SSRF, some LDAP), but leave the dominant in-band cases (SQL, XSS, most template) untouched.
- A07partialblocks — A.8.20 explicitly lists authenticating systems on the network, restricting/filtering connections (e.g. firewalls), detecting/restricting/authenticating device connections, hardening devices, segregating admin channels, disabling vulnerable protocols, and virtualized network separation; these directly block many authentication bypass, credential stuffing, session hijacking, and weak reset vectors at the network layer, but leave application-layer auth flaws (e.g. logic errors, weak password policies) untouched.
Control IDs, short titles and the structured attribute table (control type, CIA properties, cybersecurity-concept, operational capability, security domain) are facts from ISO/IEC 27001:2022 Annex A / ISO/IEC 27002:2022. The full implementation guidance prose lives in ISO/IEC 27002:2022 — not reproduced here. Cross-walks to NIST 800-53, NIST CSF 2.0, OWASP ASVS, CWE, MITRE ATT&CK and OWASP Web Top 10 are our own analysis (mostly llm_verified), not an ISO, NIST, MITRE or OWASP product — how ours compare.