Out-of-bounds Read Affecting nvidia-pytriton package, versions [0,]
Threat Intelligence
The probability is the direct output of the EPSS model, and conveys an overall sense of the threat of exploitation in the wild. The percentile measures the EPSS probability relative to all known EPSS scores. Note: This data is updated daily, relying on the latest available EPSS model version. Check out the EPSS documentation for more details.
Do your applications use this vulnerable package?
In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.
Test your applications- Snyk IDSNYK-PYTHON-NVIDIAPYTRITON-13645466
- published21 Oct 2025
- disclosed6 Aug 2025
- creditTrend Micro ZDI
Introduced: 6 Aug 2025
CVE-2025-23333 (opens in a new tab)How to fix?
There is no fixed version for nvidia-pytriton.
Overview
nvidia-pytriton is a PyTriton - Flask/FastAPI-like interface to simplify Triton's deployment in Python environments.
Affected versions of this package are vulnerable to Out-of-bounds Read via the MemoryShm::byte_size in the embedded Python backend. An attacker can cause an information disclosure by overwriting MemoryShm::byte_size data with a very large value. Successful exploitation of this vulnerability will cause the identity model to read a large chunk of sensitive data (e.g., glibc.so) as an input tensor, copy it to the output tensor, and send it back to the client.
Note: This vulnerability is only exploitable when using the default bundled Python backend /pytriton/tritonserver/backends/python/libtriton_python.so.
Workaround
It is possible to update the Python backend to a patched version independently of PyTriton; See Triton Inference Server Backend and Building binaries from source.