flash-attention contains an insecure deserialization vulnerability in its checkpoint loading mechanism

published 2026-05-11 5 mo ago modified 2026-07-14 3 mo ago also GHSA-7g5w-pq96-8c5w , PYSEC-2026-2477 , EUVD-2026-29100

cvss: 7.3; version 3.1, CNA
cisa kev: Not listed; not in CISA's catalog
epss: 0.2%; EPSS 0.2% chance of exploitation in 30 days, percentile 12%
public exploits: 0; none known

The flash-attention training framework thru commit e724e2588cbe754beb97cf7c011b5e7e34119e62 (2025-13-04) contains an insecure deserialization vulnerability (CWE-502) in its checkpoint loading mechanism. The load_checkpoint() function in checkpoint.py and the checkpoint loading code in eval.py use torch.load() without enabling the security-restrictive weights_only=True parameter. This allows the deserialization of arbitrary Python objects via the pickle module. An attacker can exploit this by providing a maliciously crafted checkpoint file. When a victim loads this checkpoint during model warmstarting or evaluation, arbitrary code is executed on the victim's system.

CISA SSVC

Exploitation: none
Automatable: no
Technical impact: partial
Assessed: 5 mo ago (2026-05-12)

Exploit prediction

0.22% chance of exploitation in the next 30 days, higher than 12% of all scored CVEs (FIRST EPSS, 2026-11-08).

CVSS scores

assigned by	version	score	vector
CISA ADP	3.1	7.3	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L
CNA	3.1	7.3	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L
GitHub	3.1	7.3	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L
OSV	3.1	7.3	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L

Affected packages

ecosystem	package	affected	fixed in
pypi	flash-attn	<=2.8.3	no fix listed

Classification

Weakness

CWE-502 Deserialization of Untrusted Data
CWE-94 Improper Control of Generation of Code

References

github.com /Dao-AILab/flash-attention
package
notion.so /CVE-2026-31253-35d1e1393188813f9e77e2038104bc49
web
nvd.nist.gov /vuln/detail/CVE-2026-31253
advisory
pypi.org /project/flash-attn
package
github.com /advisories/GHSA-7g5w-pq96-8c5w
advisory

Sources

Fencer Vulns keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.

Summary