CVE-2025-55558: PyTorch Inductor compiler buffer

CISO Take

PyTorch v2.7.0's Inductor compiler (torch.compile) has a buffer overflow triggered by specific layer combinations (Conv2d + hardshrink + view/mv), enabling unauthenticated remote attackers to crash inference services with no privileges required. Any production deployment using torch.compile with these layers is exposed. Immediate action: audit workloads for torch.compile usage, apply the patch from PR #151887, or disable Inductor as a workaround until a patched release ships.

What is the risk?

High severity (CVSS 7.5) with low attack complexity and no authentication required makes this a credible threat to ML inference APIs. Network-accessible attack surface means any externally exposed PyTorch inference endpoint running v2.7.0 with torch.compile enabled is vulnerable. Impact is limited to availability (DoS)—not data exfiltration—but ML inference service outages can cascade into production incidents and SLA breaches. EPSS not yet available; no known active exploitation at time of publication.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
PyTorch	pip	—	No patch
100.9K OpenSSF 6.4 22.7K dependents Pushed 3d ago 11% patched ~216d to patch Full package profile →

Do you use PyTorch? You're affected.

How severe is it?

CVSS 3.1

7.5 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 34% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○ CISA SSVC: Public PoC

○ Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

What should I do?

1 step

1) Patch: Apply fix from https://github.com/pytorch/pytorch/pull/151887 once a patched release is available. 2) Workaround: Disable torch.compile (Inductor) for models using Conv2d + hardshrink + view/mv combinations—run in eager mode (torch.compile disabled by default in older PyTorch versions). 3) Audit: Grep codebases and model registries for torch.compile usage combined with Conv2d and hardshrink activations. 4) Monitor: Alert on abnormal process crashes, OOM events, or container restarts in PyTorch inference workers. 5) Harden: Deploy model serving behind rate limiters and input validation layers to reduce DoS attack surface.

What does CISA's SSVC say?

Decision Track*

Exploitation poc

Automatable No

Technical Impact partial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

DoS Framework Inference AML.T0010.001 - AI Software AML.T0029 - Denial of AI Service AML.T0040 - AI Model Inference API Access AML.T0049 - Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, robustness and cybersecurity

ISO 42001

8.4 - AI system risk management

NIST AI RMF

MANAGE 2.2 - Mechanisms for incident response and recovery

OWASP LLM Top 10

LLM04 - Model Denial of Service

Frequently Asked Questions

What is CVE-2025-55558?

PyTorch v2.7.0's Inductor compiler (torch.compile) has a buffer overflow triggered by specific layer combinations (Conv2d + hardshrink + view/mv), enabling unauthenticated remote attackers to crash inference services with no privileges required. Any production deployment using torch.compile with these layers is exposed. Immediate action: audit workloads for torch.compile usage, apply the patch from PR #151887, or disable Inductor as a workaround until a patched release ships.

Is CVE-2025-55558 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2025-55558, increasing the risk of exploitation.

How to fix CVE-2025-55558?

1) Patch: Apply fix from https://github.com/pytorch/pytorch/pull/151887 once a patched release is available. 2) Workaround: Disable torch.compile (Inductor) for models using Conv2d + hardshrink + view/mv combinations—run in eager mode (torch.compile disabled by default in older PyTorch versions). 3) Audit: Grep codebases and model registries for torch.compile usage combined with Conv2d and hardshrink activations. 4) Monitor: Alert on abnormal process crashes, OOM events, or container restarts in PyTorch inference workers. 5) Harden: Deploy model serving behind rate limiters and input validation layers to reduce DoS attack surface.

What systems are affected by CVE-2025-55558?

This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, inference APIs.

What is the CVSS score for CVE-2025-55558?

CVE-2025-55558 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.42%.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesinference APIs

MITRE ATLAS Techniques

AML.T0010.001 AI Software

AML.T0029 Denial of AI Service

AML.T0040 AI Model Inference API Access

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: 8.4

NIST AI RMF: MANAGE 2.2

OWASP LLM Top 10: LLM04

What are the technical details?

Original Advisory

A buffer overflow occurs in pytorch v2.7.0 when a PyTorch model consists of torch.nn.Conv2d, torch.nn.functional.hardshrink, and torch.Tensor.view-torch.mv() and is compiled by Inductor, leading to a Denial of Service (DoS).

Exploitation Scenario

An adversary identifies a public-facing ML inference API (e.g., a computer vision or image classification service) running PyTorch 2.7.0 with torch.compile enabled. By fingerprinting the service or reviewing public model cards, they determine the model architecture includes Conv2d and hardshrink layers. The attacker submits crafted inference requests designed to trigger the specific Conv2d + hardshrink + view/mv code path in the Inductor-compiled model. The buffer overflow crashes the inference worker process. With no authentication required and low complexity, this can be scripted for sustained DoS, rendering the AI service unavailable. In Kubernetes or ECS environments, repeated crashes may exhaust pod restart budgets and take down the entire inference tier.

Weaknesses (CWE)

CWE-400 Uncontrolled Resource Consumption

CWE-400 — Uncontrolled Resource Consumption: The product does not properly control the allocation and maintenance of a limited resource.

[Architecture and Design] Design throttling mechanisms into the system architecture. The best protection is to limit the amount of resources that an unauthorized user can cause to be expended. A strong authentication and access control model will help prevent such attacks from occurring in the first place. The login application should be protected against DoS attacks as much as possible. Limiting the database access, perhaps by caching result sets, can help minimize the resources expended. To further limit the potential for a DoS attack, consider tracking the rate of requests received from users and blocking requests that exceed a defined rate threshold.
[Architecture and Design] Mitigation of resource exhaustion attacks requires that the target system either: The first of these solutions is an issue in itself though, since it may allow attackers to prevent the use of the system by a particular valid user. If the attacker impersonates the valid user, they may be able to prevent the user from accessing the server in question. The second solution is simply difficult to effectively institute -- and even when properly done, it does not provide a full solution. It simply makes the attack require more resources on the part of the attacker. recognizes the attack and denies that user further access for a given amount of time, or uniformly throttles all requests in order to make it more difficult to consume resources more quickly than they can again be freed.

Source: MITRE CWE corpus.