Cyber Resilience

CVE-2025-65890

HighPublic PoCDDoS

Published: 28 January 2026

Published
28 January 2026
Modified
03 February 2026
KEV Added
Patch
CVSS Score v3.1 7.5 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
EPSS Score 0.0005 16.9th percentile
Risk Priority 15 60% EPSS · 20% KEV · 20% CVSS

Summary

CVE-2025-65890 is a high-severity Uncontrolled Resource Consumption (CWE-400) vulnerability in Oneflow Oneflow. Its CVSS base score is 7.5 (High).

Operationally, exploitation aligns with the MITRE ATT&CK technique Exploit Public-Facing Application (T1190); ranked at the 16.9th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog; a public proof-of-concept is referenced.

The strongest mitigations our analysis identified are NIST 800-53 SC-5 (Denial-of-service Protection) and SI-10 (Information Input Validation).

Deeper analysis

CVE-2025-65890 is a device-ID validation flaw present in OneFlow version 0.9.0. This vulnerability affects the flow.cuda.synchronize() function, where attackers can supply an invalid or out-of-range GPU device index, leading to a Denial of Service (DoS) condition. The issue is classified under CWE-400 (Uncontrolled Resource Consumption) and carries a CVSS v3.1 base score of 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H), highlighting its potential for high availability impact without compromising confidentiality or integrity.

Remote attackers can exploit this vulnerability over the network with low attack complexity, requiring no privileges, authentication, or user interaction. By invoking flow.cuda.synchronize() with a malformed GPU device index, adversaries can disrupt affected OneFlow processes, potentially causing resource exhaustion or crashes on systems utilizing CUDA for GPU acceleration.

For mitigation guidance and patches, security practitioners should consult the vendor references, including the OneFlow website at http://oneflow.com, the project repository at https://github.com/Oneflow-Inc/oneflow, and the dedicated issue tracker at https://github.com/Oneflow-Inc/oneflow/issues/10662. Additional context is available from related GitHub profiles such as https://github.com/Daisy2ang.

EU & UK References

Vulnerability details

A device-ID validation flaw in OneFlow v0.9.0 allows attackers to cause a Denial of Service (DoS) by calling flow.cuda.synchronize() with an invalid or out-of-range GPU device index.

CWE(s)

Related Threats

MITRE ATT&CK Enterprise TechniquesAI

T1190 Exploit Public-Facing Application Initial Access
Adversaries may attempt to exploit a weakness in an Internet-facing host or system to initially access a network.
T1499.004 Application or System Exploitation Impact
Adversaries may exploit software vulnerabilities that can cause an application or system to crash and deny availability to users.
Why these techniques?

Vulnerability allows remote unauthenticated exploitation of a public-facing ML framework service (T1190) to trigger application/system DoS via invalid input leading to resource exhaustion or crash (T1499.004).

Confidence: HIGH · MITRE ATT&CK Enterprise v18.1

CVEs Like This One

CVE-2025-65889Same product: Oneflow Oneflow
CVE-2025-65886Same product: Oneflow Oneflow
CVE-2025-65888Same product: Oneflow Oneflow
CVE-2025-65891Same product: Oneflow Oneflow
CVE-2025-71000Same product: Oneflow Oneflow
CVE-2025-70999Same product: Oneflow Oneflow
CVE-2025-71003Same product: Oneflow Oneflow
CVE-2025-71007Same product: Oneflow Oneflow
CVE-2026-39304Shared CWE-400
CVE-2025-27669Shared CWE-400

Affected Assets

oneflow
oneflow
0.9.0

Mitigating Controls

Mitigating Controls (NIST 800-53 r5) AI

prevent

Directly addresses the device-ID validation flaw by requiring validation of GPU device indices supplied to flow.cuda.synchronize() to prevent invalid inputs causing DoS.

prevent

Mitigates the vulnerability through timely identification, reporting, and correction of the input validation flaw in OneFlow v0.9.0.

preventdetect

Protects against the DoS impact from invalid GPU device indices by limiting effects of resource exhaustion attacks on CUDA processes.

References