CVE-2024-34359 in the critical feed json
llama-cpp-python vulnerable to Remote Code Execution by Server-Side Template Injection in Model Metadata
published 2024-05-10 2 y ago modified 2026-09-10 1 mo ago also GHSA-56xg-wfcc-g829 , PYSEC-2026-392 , EUVD-2024-1433
- cvss
- 9.7
- version 3.1, CNA
- cisa kev
- Not listed
- not in CISA's catalog
- epss
- 26.0%
- EPSS 26.0% chance of exploitation in 30 days, percentile 97%
- public exploits
- 0
- none known
Summary
llama-cpp-python is the Python bindings for llama.cpp. `llama-cpp-python` depends on class `Llama` in `llama.py` to load `.gguf` llama.cpp or Latency Machine Learning Models. The `__init__` constructor built in the `Llama` takes several parameters to configure the loading and running of the model. Other than `NUMA, LoRa settings`, `loading tokenizers,` and `hardware settings`, `__init__` also loads the `chat template` from targeted `.gguf` 's Metadata and furtherly parses it to `llama_chat_format.Jinja2ChatFormatter.to_chat_handler()` to construct the `self.chat_handler` for this model. Nevertheless, `Jinja2ChatFormatter` parse the `chat template` within the Metadate with sandbox-less `jinja2.Environment`, which is furthermore rendered in `__call__` to construct the `prompt` of interaction. This allows `jinja2` Server Side Template Injection which leads to remote code execution by a carefully constructed payload.
Patch / fix
CISA SSVC
- Exploitation
- Proof of concept is public
- Automatable
- no
- Technical impact
- total
- Assessed
- 2 y ago
Exploit prediction
26.00% chance of exploitation in the next 30 days, higher than 97% of all scored CVEs (FIRST EPSS, 2026-11-08).
CVSS scores
| assigned by | version | score | vector |
|---|---|---|---|
| CNA | 3.1 | 9.7 | CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H |
| OSV | 3.1 | 9.6 | CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H |
| GitHub | 3.1 | 9.6 | CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H |
Affected packages
| ecosystem | package | affected | fixed in |
|---|---|---|---|
| pypi | llama-cpp-python | >=0.2.30, <0.2.72 |
0.2.72 |
Affected products
- abetlen llama-cpp-python
Classification
- Weakness
-
- CWE-76 Improper Neutralization of Equivalent Special Elements
References
-
github.com
/abetlen/llama-cpp-python/commit/b454f40a9a1787b2b5659cd2cb00819d983185df
web
-
github.com
/abetlen/llama-cpp-python/security/advisories/GHSA-56xg-wfcc-g829
web
-
nvd.nist.gov
/vuln/detail/CVE-2024-34359
advisory
-
github.com
/abetlen/llama-cpp-python
package
-
pypi.org
/project/llama-cpp-python
package
-
github.com
/advisories/GHSA-56xg-wfcc-g829
advisory
Sources
- CVE List CVE-2024-34359
- GitHub Advisory GHSA-56xg-wfcc-g829
- NVD CVE-2024-34359
- OSV PYSEC-2026-392
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.