CVE-2026-94093: Deserialization in DLR-RM stable-baselines3
CVE-2026-94093 is a medium severity vulnerability in DLR-RM stable-baselines3 up to version 2.9.0 involving unsafe deserialization in the PPO.load, load_replay_buffer, and VecNormalize.load functions. This flaw allows remote attackers to exploit insecure pickle deserialization, potentially leading to code execution or other impacts. Although a hardening fix was introduced in version 2.9.0, it was reverted in later development branches, leaving the core load APIs exploitable until a forthcoming fix is released. The vulnerability has been publicly disclosed but no known exploits are currently active in the wild.
AI Analysis
Technical Summary
This vulnerability affects the deserialization process in stable-baselines3's core loading functions (PPO.load, load_replay_buffer, VecNormalize.load) in versions up to 2.9.0. The issue arises from unsafe pickle deserialization, which can be manipulated remotely. A partial mitigation was introduced in v2.9.0 by hardening the PyTorch tensor load path with weights_only=True, but this was reverted on the master branch to maintain compatibility with PyTorch 1.13. Previous fixes only gated the Hugging Face Hub loader in a separate package and did not address the core APIs. The vulnerability remains exploitable until the pending hardening fix (PR #2264) is merged and released.
Potential Impact
Exploitation of this vulnerability could allow remote attackers to execute arbitrary code or perform unauthorized actions by manipulating deserialized data. The CVSS 4.0 score of 5.3 reflects a medium severity with network attack vector, low complexity, no privileges required, and user interaction needed. Confidentiality, integrity, and availability impacts are rated low to limited.
Mitigation Recommendations
A patch is available and should be applied once the outstanding hardening fix (PR #2264) is released. Until then, users should avoid loading untrusted serialized data with the affected APIs. The vendor manages remediation for this cloud-hosted service. Check the vendor advisory for the latest status and apply official fixes promptly.
CVE-2026-94093: Deserialization in DLR-RM stable-baselines3
Description
CVE-2026-94093 is a medium severity vulnerability in DLR-RM stable-baselines3 up to version 2.9.0 involving unsafe deserialization in the PPO.load, load_replay_buffer, and VecNormalize.load functions. This flaw allows remote attackers to exploit insecure pickle deserialization, potentially leading to code execution or other impacts. Although a hardening fix was introduced in version 2.9.0, it was reverted in later development branches, leaving the core load APIs exploitable until a forthcoming fix is released. The vulnerability has been publicly disclosed but no known exploits are currently active in the wild.
CVSS v4.0
Score 5.3medium
Affected software
DLR-RM
stable-baselines3
cpe:2.3:a:dlr-rm:stable-baselines3:*:*:*:*:*:*:*:*AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
This vulnerability affects the deserialization process in stable-baselines3's core loading functions (PPO.load, load_replay_buffer, VecNormalize.load) in versions up to 2.9.0. The issue arises from unsafe pickle deserialization, which can be manipulated remotely. A partial mitigation was introduced in v2.9.0 by hardening the PyTorch tensor load path with weights_only=True, but this was reverted on the master branch to maintain compatibility with PyTorch 1.13. Previous fixes only gated the Hugging Face Hub loader in a separate package and did not address the core APIs. The vulnerability remains exploitable until the pending hardening fix (PR #2264) is merged and released.
Potential Impact
Exploitation of this vulnerability could allow remote attackers to execute arbitrary code or perform unauthorized actions by manipulating deserialized data. The CVSS 4.0 score of 5.3 reflects a medium severity with network attack vector, low complexity, no privileges required, and user interaction needed. Confidentiality, integrity, and availability impacts are rated low to limited.
Mitigation Recommendations
A patch is available and should be applied once the outstanding hardening fix (PR #2264) is released. Until then, users should avoid loading untrusted serialized data with the affected APIs. The vendor manages remediation for this cloud-hosted service. Check the vendor advisory for the latest status and apply official fixes promptly.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- VulDB
- Date Reserved
- 2026-09-20T08:39:03.152Z
- Cvss Version
- 4.0
- State
- PUBLISHED
- Is Cloud Service
- true
Threat ID: 6ab065f455bf5e2cf5a51ffd
Added to database: 09/20/2026, 23:02:12 UTC
Last enriched: 09/20/2026, 23:16:31 UTC
Last updated: 09/20/2026, 23:21:26 UTC
Views: 5
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.