CVE-2023-25661
📋 TL;DR
This vulnerability in TensorFlow allows an attacker with input privileges to provide malicious data to the Convolution3DTranspose function, causing a crash and denial of service. It affects TensorFlow versions before 2.11.1 when models use this common neural network layer. Organizations using vulnerable TensorFlow versions in ML applications or cloud services are at risk.
💻 Affected Systems
- TensorFlow
📦 What is this software?
⚠️ Risk & Real-World Impact
Worst Case
Complete denial of service for ML cloud services or applications using vulnerable TensorFlow models, disrupting business operations.
Likely Case
Application crashes requiring restart when malicious input reaches Convolution3DTranspose layers in deployed models.
If Mitigated
Limited impact with input validation and proper access controls, though crash still possible with valid malicious input.
🎯 Exploit Status
Proof of concept exists in advisory, requires input access to Convolution3DTranspose function.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 2.11.1
Vendor Advisory: https://github.com/tensorflow/tensorflow/security/advisories/GHSA-fxgc-95xx-grvq
Restart Required: Yes
Instructions:
1. Upgrade TensorFlow to version 2.11.1 or later using pip: 'pip install --upgrade tensorflow==2.11.1' 2. Restart all services using TensorFlow 3. Rebuild and redeploy any affected ML models
🔧 Temporary Workarounds
No known workarounds
allThe advisory states there are no known workarounds for this vulnerability
🧯 If You Can't Patch
- Implement strict input validation and sanitization for all inputs to TensorFlow models
- Restrict access to ML inference endpoints to trusted users only
🔍 How to Verify
Check if Vulnerable:
Check TensorFlow version: 'python -c "import tensorflow as tf; print(tf.__version__)"' - if version < 2.11.1 and using Convolution3DTranspose, vulnerable.Check Version:
python -c "import tensorflow as tf; print(tf.__version__)"Verify Fix Applied:
Verify TensorFlow version is 2.11.1 or higher and test Convolution3DTranspose with previously crashing inputs.📡 Detection & Monitoring
Log Indicators:
- TensorFlow crash logs with 'Check Failed' errors
- Application crashes when processing Convolution3DTranspose operations
Network Indicators:
- Unusual patterns of requests to ML inference endpoints followed by service crashes
SIEM Query:
source="tensorflow" AND ("Check Failed" OR "Convolution3DTranspose" AND error)🔗 References
- https://github.com/tensorflow/tensorflow/commit/948fe6369a5711d4b4568ea9bbf6015c6dfb77e2
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-fxgc-95xx-grvq
- https://github.com/tensorflow/tensorflow/commit/948fe6369a5711d4b4568ea9bbf6015c6dfb77e2
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-fxgc-95xx-grvq
