ReDoS Vulnerability in Hugging Face Transformers Library

CVE-2025-6051

What is CVE-2025-6051?

A vulnerability has been identified in the Hugging Face Transformers library related to excessive CPU consumption due to improper handling of numeric strings in the 'normalize_numbers()' method of the 'EnglishNormalizer' class. Attackers can exploit this vulnerability by supplying crafted input strings containing long sequences of digits, leading to denial-of-service conditions. This can disrupt critical text-to-speech and number normalization processes, potentially exhausting system resources and compromising API stability. The issue has been resolved in version 4.53.0.

References