CVE-2025-2148
Published: 10 March 2025
Description
Adversaries may exploit software vulnerabilities that can cause an application or system to crash and deny availability to users.
Security Summary
CVE-2025-2148 is a critical vulnerability in PyTorch version 2.6.0+cu124, affecting the function torch.ops.profiler._call_end_callbacks_on_jit_fut within the Tuple Handler component. The issue stems from manipulation of the argument None, resulting in memory corruption classified under CWE-119 (Improper Restriction of Operations within the Bounds of a Memory Buffer). It carries a CVSS v3.1 base score of 5.0 (AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:L/A:L) and was published on 2025-03-10.
Remote attackers with no privileges can exploit this vulnerability, though it requires high attack complexity and user interaction. Successful exploitation leads to limited impacts, including low confidentiality, integrity, and availability effects through memory corruption.
Advisories and further details are available in the referenced sources, including PyTorch GitHub issues at https://github.com/pytorch/pytorch/issues/147722 and VulDB entries at https://vuldb.com/?ctiid.299059, https://vuldb.com/?id.299059, and https://vuldb.com/?submit.505959, which security practitioners should consult for patch information and mitigation guidance.
Details
- CWE(s)
Affected Products
AI Security Analysis
- AI Category
- Deep Learning Frameworks
- Risk Domain
- Other ATLAS/OWASP Terms
- OWASP Top 10 for LLMs 2025
- None mapped
- MITRE ATLAS Techniques
- None mapped
- Classification Reason
- PyTorch is a widely used deep learning framework, and the vulnerability affects a core component (torch.ops.profiler._call_end_callbacks_on_jit_fut) involving tuple handling and JIT operations, confirming it as AI-related in the deep learning domain.
MITRE ATT&CK Enterprise Techniques
Why these techniques?
The remote memory corruption vulnerability (CWE-119) in PyTorch's profiler function causes a segmentation fault/crash upon manipulation with a None argument, enabling application-level denial of service via exploitation.