Skip to content
Fencer Vulns
low 3.4

CVE-2024-3568 json

Arbitrary Code Execution via Deserialization in huggingface/transformers

published 2024-04-10 2 y ago modified 2026-07-07 4 mo ago also GHSA-37q5-v5qm-c9v8 , PYSEC-2026-1978 , EUVD-2024-1077

cvss
3.4
version 3.0, CNA
cisa kev
Not listed
not in CISA's catalog
epss
2.1%
EPSS 2.1% chance of exploitation in 30 days, percentile 81%
public exploits
0
none known

Summary

The huggingface/transformers library is vulnerable to arbitrary code execution through deserialization of untrusted data within the `load_repo_checkpoint()` function of the `TFPreTrainedModel()` class. Attackers can execute arbitrary code and commands by crafting a malicious serialized payload, exploiting the use of `pickle.load()` on data from potentially untrusted sources. This vulnerability allows for remote code execution (RCE) by deceiving victims into loading a seemingly harmless checkpoint during a normal training process, thereby enabling attackers to execute arbitrary code on the targeted machine.

Patch / fix

  • pypi transformers 4.38.0
  • huggingface huggingface/transformers 4.38

CISA SSVC

Exploitation
none
Automatable
no
Technical impact
partial
Assessed
2 y ago

Exploit prediction

2.08% chance of exploitation in the next 30 days, higher than 81% of all scored CVEs (FIRST EPSS, 2026-11-08).

Exploits and checks

  • huntr.com https://huntr.com/bounties/b3c36992-5264-4d7f-9906-a996efafba8f referenced PoC, tagged "exploit" by a source, unverified

CVSS scores

assigned by version score vector
CNA 3.0 3.4 CVSS:3.0/AV:N/AC:H/PR:N/UI:R/S:C/C:N/I:N/A:L
NVD 3.1 9.6 CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H
GitHub 3.0 3.4 CVSS:3.0/AV:N/AC:H/PR:N/UI:R/S:C/C:N/I:N/A:L
OSV 3.0 3.4 CVSS:3.0/AV:N/AC:H/PR:N/UI:R/S:C/C:N/I:N/A:L

Affected packages

ecosystem package affected fixed in
pypi transformers
<4.38.0
4.38.0

Affected products

  • huggingface huggingface/transformers

Classification

Weakness
  • CWE-502 Deserialization of Untrusted Data

References

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.