CVE-2025-3264
📋 TL;DR
A Regular Expression Denial of Service (ReDoS) vulnerability in Hugging Face Transformers library allows attackers to cause excessive CPU consumption by providing specially crafted input strings to the `get_imports()` function. This affects users running Transformers version 4.49.0 for model serving, code loading, or development pipelines. The vulnerability can lead to service disruption and resource exhaustion.
💻 Affected Systems
- Hugging Face Transformers
📦 What is this software?
Transformers by Huggingface
⚠️ Risk & Real-World Impact
Worst Case
Complete denial of service in model serving environments, disrupting AI inference services and potentially causing cascading failures in dependent systems.
Likely Case
Temporary service degradation or disruption during code loading operations, particularly affecting development pipelines and model deployment workflows.
If Mitigated
Minimal impact with proper input validation, rate limiting, and updated library versions.
🎯 Exploit Status
Proof of concept available in public bounty reports. Exploitation requires ability to provide input to the vulnerable function, which typically requires some level of access or API interaction.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 4.51.0
Vendor Advisory: https://github.com/huggingface/transformers/commit/0720e206c6ba28887e4d60ef60a6a089f6c1cc76
Restart Required: No
Instructions:
1. Update Hugging Face Transformers using pip: `pip install transformers>=4.51.0` 2. Verify the update with `pip show transformers` 3. Test that your application functions correctly with the new version.
🔧 Temporary Workarounds
Input Validation and Sanitization
allImplement strict input validation for any user-provided code or strings that might reach the `get_imports()` function.
Version Downgrade
allTemporarily downgrade to version 4.48.0 or earlier until patching is possible.
pip install transformers==4.48.0
🧯 If You Can't Patch
- Implement rate limiting and timeout mechanisms for code processing operations
- Isolate Transformers processing to dedicated containers with CPU limits and monitoring
🔍 How to Verify
Check if Vulnerable:
Check Transformers version with: `python -c "import transformers; print(transformers.__version__)"` and verify if it equals '4.49.0'Check Version:
python -c "import transformers; print(transformers.__version__)"Verify Fix Applied:
After updating, verify version is 4.51.0 or higher with same command📡 Detection & Monitoring
Log Indicators:
- Unusually high CPU usage in Transformers processes
- Timeout errors in code loading operations
- Repeated failed import attempts
Network Indicators:
- Increased latency in model serving endpoints
- Timeout responses from AI inference services
SIEM Query:
process.name:"python" AND process.args:"transformers" AND cpu.usage > 90%