Executive brief
OneFlow is a deep learning framework used for building and training artificial intelligence models. A vulnerability in the way it handles mathematical operations between different data structures can cause the software to crash. An attacker could exploit this to disrupt AI services or research environments, leading to a denial of service.
Technical details
A shape mismatch vulnerability exists in OneFlow v0.9.0 when performing operations between tensors and other data structures, specifically involving the 'flow.eye' and 'diag' functions. The root cause is a failure to properly validate or handle inconsistent dimensions between operands, leading to a segmentation fault and core dump (CWE-400). An attacker can trigger this crash by supplying crafted tensor shapes to the affected functions. The vulnerability is reachable via network if the framework is exposed through an API that processes user-supplied tensor dimensions. As of the advisory, this is identified in version 0.9.0.
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