Arbitrary code execution via unsafe TorchModuleWrapper deserialization
Published Oct 17, 2025 · Updated Oct 17, 2025
Deserialization in Keras 3.11.0 through 3.11.2 allows remote attackers to execute arbitrary code through a crafted model file. The TorchModuleWrapper.from_config method base64-decodes the model's module field and passes it to torch.load with weights_only=False even in safe mode. Loading an attacker-controlled local or remote .keras model is the gate, and the unpickled payload runs with the loading process's privileges.
Summary
What happened
Deserialization in Keras 3.11.0 through 3.11.2 allows remote attackers to execute arbitrary code through a crafted model file. The TorchModuleWrapper.from_config method base64-decodes the model's module field and passes it to torch.load with weights_only=False even in safe mode. Loading an attacker-controlled local or remote .keras model is the gate, and the unpickled payload runs with the loading process's privileges.
The record
- CVE
- CVE-2025-49655
- Published
- Oct 17, 2025
- Updated
- Oct 17, 2025
- Vendor
- Keras
- Product
- Keras
- Classifications
- CWE-502, T1203
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- Oct 17, 2025CVE publishedPublication date reported by the CVE source.
- Oct 17, 2025Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=3.11.0 <3.11.3
What conditions does exploitation require?
What is affected?
Published CVSS scores
CVSS describes severity. EPSS estimates exploitation probability.
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- Crafted TorchModuleWrapper model proof of conceptproof of concept · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo