CVE-2024-48063
Published: 29 October 2024
Summary
CVE-2024-48063 is a critical-severity Deserialization of Untrusted Data (CWE-502) vulnerability in Linuxfoundation Pytorch. Its CVSS base score is 9.8 (Critical).
Operationally, exploitation aligns with the MITRE ATT&CK technique Exploitation of Remote Services (T1210); ranked in the top 3.7% of CVEs by exploit likelihood; it is not currently listed in the CISA KEV catalog; a public proof-of-concept is referenced.
This vulnerability is AI-related — categorised as Deep Learning Frameworks; in the Protocol-Specific Risks risk domain; MITRE ATLAS techniques in scope: Hardware (AML.T0010.000), Infer Training Data Membership (AML.T0024.000), Financial Harm (AML.T0048.000).
Deeper analysis
In PyTorch versions up to and including 2.4.1, the RemoteModule component contains a deserialization vulnerability tracked as CVE-2024-48063 and assigned CWE-502. The issue enables remote code execution through unsafe handling of serialized data and carries a CVSS 3.1 score of 9.8. Multiple parties have disputed the classification, stating that the behavior is intentional within PyTorch's distributed computing framework rather than a flaw.
An unauthenticated remote attacker can supply malicious serialized payloads to a RemoteModule instance over the network, achieving arbitrary code execution with full confidentiality, integrity, and availability impact. Exploitation requires no user interaction and targets the distributed RPC features that accept and deserialize objects from remote participants.
PyTorch's security policy on distributed features and the associated GitHub issue advise that users should restrict distributed RPC usage to trusted environments and follow documented secure deployment practices for RemoteModule and related components. No separate patch is indicated beyond these usage guidelines.
The vulnerability affects an AI/ML framework and shows an EPSS score that rose from a low baseline to a peak of 0.3003, indicating emerging exploitation interest after disclosure.
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2024-42923
Vulnerability details
In PyTorch <=2.4.1, the RemoteModule has Deserialization RCE. NOTE: this is disputed by multiple parties because this is intended behavior in PyTorch distributed computing.
- CWE(s)
AI Security AnalysisAI
- AI Category
- Deep Learning Frameworks
- Risk Domain
- Protocol-Specific Risks
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- PyTorch is a core deep learning framework, and the vulnerability affects its distributed RPC framework used for distributed machine learning tasks across nodes.
Related Threats
MITRE ATT&CK Enterprise TechniquesAI
Why these techniques?
The deserialization RCE vulnerability in PyTorch's distributed RPC framework (rpc.remote) enables remote attackers to execute arbitrary code by sending malicious payloads over the network, directly facilitating T1210: Exploitation of Remote Services.
MITRE ATLAS TechniquesAI
MITRE ATLAS techniques
Affected Assets
Mitigating Controls
Likely Mitigating Controls AI
Per-CVE control mapping for this CVE has not run yet; the list below is derived from the weakness types (CWEs) cited in the NVD entry.
Penetration testing supplies malicious serialized objects, detecting unsafe deserialization and supporting corrective actions.
Evaluation of untrusted data handling (deserialization testing) reveals unsafe processing, which the required remediation process addresses.
Untrusted serialized data can be deserialized and observed inside the chamber, blocking gadget-chain exploitation outside the sandbox.
Validates or rejects untrusted serialized data before deserialization occurs.
Identifies and blocks malicious code introduced through deserialization of untrusted data at system boundaries.
Integrity verification of serialized information can detect tampering before deserialization occurs.
Provenance of associated data allows detection of untrusted sources before deserialization or processing occurs.