CVE-2025-6051: Transformers ReDoS — MEDIUM

CISO Take

Upgrade Hugging Face Transformers to 4.53.0 immediately if you run any TTS or text normalization pipelines that accept external input. This ReDoS is trivially exploitable with no authentication — a single crafted string of digits can peg a CPU core and degrade or kill your inference service. The EPSS score is near zero today, but the attack pattern is trivial enough that exploitation could spike once PoC details spread from the Huntr disclosure.

What is the risk?

Effective risk is moderate-to-high for externally-exposed inference APIs despite the CVSS 5.3 rating. The CVSS vector (AV:N/AC:L/PR:N/UI:N) means any unauthenticated remote attacker can trigger it. Impact is purely availability (no confidentiality or integrity loss), but in production AI deployments, a sustained CPU exhaustion attack can translate to SLA breaches, autoscaling cost spikes, and cascading failures in multi-model pipelines. Low EPSS (0.00034) reflects limited active exploitation today, but the vulnerability is in one of the most-deployed ML libraries globally.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Transformers	pip	—	No patch
161.8K OpenSSF 6.4 8.3K dependents Pushed 4d ago 40% patched ~92d to patch Full package profile →
Transformers	pip	< 4.53.0	`4.53.0`
161.8K OpenSSF 6.4 8.3K dependents Pushed 4d ago 40% patched ~92d to patch Full package profile →

How severe is it?

CVSS 3.1

5.3 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 27% 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 Low

What should I do?

5 steps

PATCH

Upgrade transformers to >= 4.53.0 (pip install --upgrade transformers). Verify with pip show transformers.
WORKAROUND (if immediate patching is blocked): Add input validation upstream — reject or truncate strings exceeding a reasonable digit-run length (e.g., reject inputs with consecutive digit sequences > 50 chars).
DETECTION

Alert on CPU utilization spikes in inference workers correlated with text-processing requests; set process-level CPU limits (cgroups/K8s resource limits) to prevent one request from monopolizing the node.
INVENTORY

Audit all services importing transformers and identify which expose text input to the EnglishNormalizer code path.
VERIFY

Check the commit at ba8eaba9865618253f997784aa565b96206426f0 for the exact regex fix to understand the vulnerable pattern.

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.T0034 - Cost Harvesting 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

A.6.2.6 - AI system availability and resilience

NIST AI RMF

MANAGE-2.2 - Mechanisms to sustain value of deployed AI

OWASP LLM Top 10

LLM10:2025 - Unbounded Consumption

Frequently Asked Questions

What is CVE-2025-6051?

Upgrade Hugging Face Transformers to 4.53.0 immediately if you run any TTS or text normalization pipelines that accept external input. This ReDoS is trivially exploitable with no authentication — a single crafted string of digits can peg a CPU core and degrade or kill your inference service. The EPSS score is near zero today, but the attack pattern is trivial enough that exploitation could spike once PoC details spread from the Huntr disclosure.

Is CVE-2025-6051 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2025-6051, increasing the risk of exploitation.

How to fix CVE-2025-6051?

1. PATCH: Upgrade transformers to >= 4.53.0 (pip install --upgrade transformers). Verify with pip show transformers. 2. WORKAROUND (if immediate patching is blocked): Add input validation upstream — reject or truncate strings exceeding a reasonable digit-run length (e.g., reject inputs with consecutive digit sequences > 50 chars). 3. DETECTION: Alert on CPU utilization spikes in inference workers correlated with text-processing requests; set process-level CPU limits (cgroups/K8s resource limits) to prevent one request from monopolizing the node. 4. INVENTORY: Audit all services importing transformers and identify which expose text input to the EnglishNormalizer code path. 5. VERIFY: Check the commit at ba8eaba9865618253f997784aa565b96206426f0 for the exact regex fix to understand the vulnerable pattern.

What systems are affected by CVE-2025-6051?

This vulnerability affects the following AI/ML architecture patterns: TTS inference pipelines, NLP text normalization services, model serving, batch enrichment pipelines, training data preprocessing.

What is the CVSS score for CVE-2025-6051?

CVE-2025-6051 has a CVSS v3.1 base score of 5.3 (MEDIUM). The EPSS exploitation probability is 0.35%.

What is the AI security impact?

Affected AI Architectures

TTS inference pipelinesNLP text normalization servicesmodel servingbatch enrichment pipelinestraining data preprocessing

MITRE ATLAS Techniques

AML.T0010.001 AI Software

AML.T0029 Denial of AI Service

AML.T0034 Cost Harvesting

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE-2.2

OWASP LLM Top 10: LLM10:2025

What are the technical details?

Original Advisory

A Regular Expression Denial of Service (ReDoS) vulnerability was discovered in the Hugging Face Transformers library, specifically within the `normalize_numbers()` method of the `EnglishNormalizer` class. This vulnerability affects versions up to 4.52.4 and is fixed in version 4.53.0. The issue arises from the method's handling of numeric strings, which can be exploited using crafted input strings containing long sequences of digits, leading to excessive CPU consumption. This vulnerability impacts text-to-speech and number normalization tasks, potentially causing service disruption, resource exhaustion, and API vulnerabilities.

Exploitation Scenario

An adversary identifies a public-facing TTS or text-normalization API endpoint powered by Hugging Face Transformers (discoverable via response headers, error messages, or job listings). They craft a POST request with a body containing a long numeric string (e.g., a 50,000-digit number). The normalize_numbers() regex enters catastrophic backtracking, consuming 100% of a CPU core for tens of seconds per request. By sending a low-rate stream of such requests (avoiding traditional rate-limit thresholds), the attacker degrades service quality for all users. In a single-threaded or worker-pool model, this effectively takes the endpoint offline. No credentials, no prior access, no AI/ML knowledge required.

Weaknesses (CWE)

CWE-1333 Inefficient Regular Expression Complexity Primary CWE-1333 Inefficient Regular Expression Complexity

CWE-1333 — Inefficient Regular Expression Complexity: The product uses a regular expression with a worst-case computational complexity that is inefficient and possibly exponential.

[Architecture and Design] Use regular expressions that do not support backtracking, e.g. by removing nested quantifiers.
[System Configuration] Set backtracking limits in the configuration of the regular expression implementation, such as PHP's pcre.backtrack_limit. Also consider limits on execution time for the process.

Source: MITRE CWE corpus.