Hugging Face peft's LoRA-GA and CorDA initialization modules (src/peft/tuners/lora/corda.py lines ~102 and ~163, and src/peft/tuners/lora/loraga.py line ~101) call torch.load() on config-specified cache/covariance files without weights_only=True, bypassing peft's own safe-loading wrapper used elsewhere in the codebase. Because torch.load() without weights_only=True performs full pickle deserialization, loading a malicious cache or covariance file (e.g. a shared/downloaded LoRA-GA or CorDA cache) results in arbitrary code execution.
Project Subscriptions
No data.
Advisories
No advisories yet.
Fixes
Solution
No solution given by the vendor.
Workaround
No workaround given by the vendor.
References
History
Wed, 05 Aug 2026 13:15:00 +0000
| Type | Values Removed | Values Added |
|---|---|---|
| Description | Hugging Face peft's LoRA-GA and CorDA initialization modules (src/peft/tuners/lora/corda.py lines ~102 and ~163, and src/peft/tuners/lora/loraga.py line ~101) call torch.load() on config-specified cache/covariance files without weights_only=True, bypassing peft's own safe-loading wrapper used elsewhere in the codebase. Because torch.load() without weights_only=True performs full pickle deserialization, loading a malicious cache or covariance file (e.g. a shared/downloaded LoRA-GA or CorDA cache) results in arbitrary code execution. | |
| Title | peft Unsafe Deserialization via torch.load() Without weights_only in LoRA-GA and CorDA Modules | |
| Weaknesses | CWE-502 | |
| References |
| |
| Metrics |
cvssV3_1
|
Projects
Sign in to view the affected projects.
Status: PUBLISHED
Assigner: TuranSec
Published:
Updated: 2026-08-05T12:52:46.791Z
Reserved: 2026-08-05T12:23:34.968Z
Link: CVE-2026-71281
No data.
No data.
No data.
OpenCVE Enrichment
No data.
Weaknesses