Log in Start free

Vulnerability · Published Sep 20, 2026 · Assigned by VulDB

CVE-2026-94093DLR-RM stable-baselines3 save_util.py VecNormalize.load deserialization

Medium 6.3

In plain English

A medium-severity unsafe deserialization in stable-baselines3 from DLR-RM. Proof-of-concept exploit code is public. The chance it is exploited in the next 30 days is estimated at 0.5% (higher than 39% of all known vulnerabilities).

Severity (CVSS 3.1)6.3Medium
Exploit odds (30 days)0.5%top 60.6% of all CVEs
Exploited in the wildNo reportsCISA SSVC: poc
Attack toolingNoneno public tooling known

Official description

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.

Affected products (as reported)

VendorProductVersions
DLR-RM stable-baselines3 2.02.12.22.32.42.52.62.7

Weakness: CWE-502, CWE-20 · Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:L/E:P/RL:X/RC:R

References