Cyber Resilience

CVE-2025-55551

HighDDoS

Published: 25 September 2025

Published
25 September 2025
Modified
03 October 2025
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 0.0012 30.4th percentile
Risk Priority 15 60% EPSS · 20% KEV · 20% CVSS

Summary

CVE-2025-55551 is a high-severity Uncontrolled Resource Consumption (CWE-400) vulnerability in Linuxfoundation Pytorch. Its CVSS base score is 7.5 (High).

Operationally, exploitation aligns with the MITRE ATT&CK technique Application or System Exploitation (T1499.004); ranked at the 30.4th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog.

This vulnerability is AI-related — categorised as Deep Learning Frameworks; in the Not Applicable risk domain.

EU & UK References

Vulnerability details

An issue in the component torch.linalg.lu of pytorch v2.8.0 allows attackers to cause a Denial of Service (DoS) when performing a slice operation.

CWE(s)

AI Security AnalysisAI

AI Category
Deep Learning Frameworks
Risk Domain
Not Applicable
OWASP Top 10 for LLMs 2025
None mapped
Classification Reason
Matched keywords: pytorch

Related Threats

MITRE ATT&CK Enterprise TechniquesAI

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.
T1565.003 Runtime Data Manipulation Impact
Adversaries may modify systems in order to manipulate the data as it is accessed and displayed to an end user, thus threatening the integrity of the data.
Why these techniques?

Vulnerabilities in PyTorch Inductor and TensorFlow XLA compilers allow remote attackers to trigger DoS crashes (syntax/name/buffer errors, system crashes) via specific model compilations (T1499.004) and cause silent incorrect model outputs or integer overflows leading to runtime data manipulation (T1565.003).

Affected Assets

linuxfoundation
pytorch
≤ 2.8.0

Mitigating Controls

Likely Mitigating Controls AI

Per-CVE control mapping for this CVE has not run yet; the list below is derived from the weakness types (CWEs) cited in the NVD entry.

addresses: CWE-400

Limiting concurrent sessions directly prevents uncontrolled resource consumption by capping the number of active sessions per user or account.

addresses: CWE-400

Analysis identifies uncontrolled resource consumption indicative of denial-of-service or abuse attempts.

addresses: CWE-400

Contingency plan testing includes resource exhaustion scenarios to verify recovery, making it harder for attackers to sustain exploits that cause uncontrolled consumption.

addresses: CWE-400

Updated contingency plans include current procedures to detect, contain, and recover from resource exhaustion, limiting an attacker's ability to sustain impact from uncontrolled consumption.

addresses: CWE-400

Alternate site allows resumption of operations if resource exhaustion at the primary site is exploited to cause unavailability.

addresses: CWE-400

Alternate telecommunications services enable resumption of essential functions when primary services become unavailable due to uncontrolled resource consumption.

addresses: CWE-400

The team can analyze and respond to resource exhaustion incidents, reducing the impact of attacks that exploit uncontrolled consumption weaknesses.

addresses: CWE-400

Timely maintenance support and spare parts enable rapid recovery from failures induced by uncontrolled resource consumption, shortening the impact window of denial-of-service attacks.

References