Executive brief
OneFlow is an open-source deep learning framework used for training and deploying artificial intelligence models. A flaw in how the software handles GPU device identifiers allows an attacker to crash the application by providing an invalid ID. This results in a denial-of-service, potentially disrupting AI training workloads or production inference services.
Technical details
A vulnerability exists in OneFlow v0.9.0 within the `flow.cuda.get_device_capability()` function due to insufficient validation of the `device` parameter. When an invalid or out-of-bounds device ID is provided, the underlying `cudaSetDevice()` call fails, triggering a `Check failed` assertion in `cuda_util.cpp`. This leads to an immediate process crash (SIGABRT) and core dump. An attacker capable of influencing the device ID passed to this function can remotely or locally cause a denial-of-service (DoS) condition. The issue is documented in the project's GitHub issue tracker as a bug where the framework fails to gracefully handle invalid device ordinals.
Affected products
- OneFlow-Inc OneFlow 0.9.0
Timeline
- 2025-06-30: disclosed: Issue reported on GitHub repository
- 2026-01-28: advisory: CVE published to NVD