Code execution via unrestricted checkpoint deserialization
Published May 11, 2026 · Updated May 11, 2026
Insecure deserialization in flash-attention allows attackers to execute code by supplying a malicious checkpoint file for loading. The load_checkpoint() function in checkpoint.py and checkpoint loading code in eval.py call torch.load() without weights_only=True, allowing pickle to restore attacker-controlled Python objects. Execution occurs only when the crafted checkpoint is loaded for model warmstarting or evaluation and inherits the training process's privileges.
Summary
What happened
Insecure deserialization in flash-attention allows attackers to execute code by supplying a malicious checkpoint file for loading. The load_checkpoint() function in checkpoint.py and checkpoint loading code in eval.py call torch.load() without weights_only=True, allowing pickle to restore attacker-controlled Python objects. Execution occurs only when the crafted checkpoint is loaded for model warmstarting or evaluation and inherits the training process's privileges.
The record
- CVE
- CVE-2026-31253
- Published
- May 11, 2026
- Updated
- May 11, 2026
- Vendor
- Unknown vendor
- Product
- Unknown product
- Classifications
- CWE-502, CWE-94, T1204.002
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- May 11, 2026CVE publishedPublication date reported by the CVE source.
- May 11, 2026Record 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?
What conditions does exploitation require?
What is affected?
Affected products and versions are unavailable in this record.
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
No public exploit references are available in this record.
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