CVE Catalog

CVE-2026-94093

MediumCVSS 6.3
Published: Updated: Translated: NVD NIST

Exploitation Probability (EPSS)

Low risk
0.26%

18th percentile - higher than 18% of all known CVEs

Summary

A security vulnerability has been detected in DLR-RM stable-baselines3 up to version 2.9.0. It affects the functions PPO.load, load_replay_buffer, and VecNormalize.load in save_util.py. Such manipulation leads to deserialization. The attack can be launched remotely, and the exploit has been publicly disclosed. In v2.9.0 the PyTorch tensor load path is hardened (weights_only=True), but that hardening was later reverted on master via PR #1913, causing inconsistency across versions. The vulnerability remains in the core stable_baselines3 load APIs.

Risk Assessment

A remote attacker can supply a malicious pickle file, leading to arbitrary code execution on the server, potentially resulting in system compromise.

Recommendation

Avoid loading untrusted model files and implement safe deserialization methods. Monitor progress on PR #2264 and apply the fix when available.

Related vulnerabilities

Original NVD description (English source)

A security vulnerability has been detected in DLR-RM stable-baselines3 up to 2.9.0. This affects the function PPO.load/load_replay_buffer/VecNormalize.load of the file save_util.py. Such manipulation leads to deserialization. It is possible to launch the attack remotely. The exploit has been disclosed publicly and may be used. In v2.9.0 the PyTorch tensor load path is hardened (weights_only=True), but that hardening was later reverted on master via PR #1913 "Hotfix: revert loading with weights_only=True" [blocked] to fix PyTorch 1.13 compat - so even the one "safe" path is inconsistent across versions. #2281 was closed as a duplicate of #1831 since both are unsafe pickle deserialization - but #1831's fix (PR #41) only gated the Hugging Face Hub loader in the separate huggingface_sb3 package. This finding covers the core stable_baselines3 load APIs (PPO.load, load_replay_buffer, VecNormalize.load), which have no safe mode or gate and remained exploitable in v2.9.0 until the outstanding hardening (PR #2264) ships.

Vulnerability data from NVD (NIST) · CISA KEV · EPSS