CVE-2023-25674
📋 TL;DR
This CVE describes a null pointer dereference vulnerability in TensorFlow's RandomShuffle operation when XLA (Accelerated Linear Algebra) is enabled. It affects TensorFlow versions prior to 2.12.0 and 2.11.1, potentially causing denial-of-service (DoS) or application crashes. Users running TensorFlow with XLA enabled are primarily at risk.
💻 Affected Systems
- TensorFlow
📦 What is this software?
⚠️ Risk & Real-World Impact
Worst Case
Exploitation could lead to a denial-of-service (DoS) condition, crashing the TensorFlow process and disrupting machine learning workloads, possibly causing data loss or service unavailability.
Likely Case
Most probable impact is application instability or crashes when using RandomShuffle with XLA, resulting in temporary service interruptions for affected TensorFlow deployments.
If Mitigated
If proper controls like network segmentation and least privilege are in place, impact is limited to local DoS without remote code execution or data compromise.
🎯 Exploit Status
Exploitation requires XLA to be enabled and specific use of RandomShuffle; no public proof-of-concept has been disclosed as of the advisory.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: TensorFlow 2.12.0 or 2.11.1
Vendor Advisory: https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gf97-q72m-7579
Restart Required: Yes
Instructions:
1. Identify the current TensorFlow version. 2. Upgrade to TensorFlow 2.12.0 or 2.11.1 using pip: 'pip install --upgrade tensorflow==2.12.0' or 'pip install --upgrade tensorflow==2.11.1'. 3. Restart any services or applications using TensorFlow to apply the fix.
🔧 Temporary Workarounds
Disable XLA
allDisable XLA (Accelerated Linear Algebra) in TensorFlow configurations to mitigate the vulnerability, as it only affects RandomShuffle with XLA enabled.
Set environment variable: export TF_XLA_FLAGS=--tf_xla_auto_jit=0
Or configure TensorFlow to disable XLA in code: tf.config.optimizer.set_jit(False)
🧯 If You Can't Patch
- Disable XLA in TensorFlow configurations to prevent exploitation of the vulnerability.
- Implement network segmentation and restrict access to TensorFlow services to reduce attack surface.
🔍 How to Verify
Check if Vulnerable:
Check TensorFlow version and XLA status: run 'python -c "import tensorflow as tf; print(tf.__version__); print(tf.config.optimizer.get_jit())"' and verify if version is prior to 2.12.0 or 2.11.1 with XLA enabled.Check Version:
python -c "import tensorflow as tf; print(tf.__version__)"Verify Fix Applied:
After patching, run the same command to confirm TensorFlow version is 2.12.0 or 2.11.1 and test RandomShuffle operations with XLA to ensure no crashes occur.📡 Detection & Monitoring
Log Indicators:
- Look for application crash logs or error messages related to TensorFlow RandomShuffle operations, especially with null pointer references or segmentation faults.
Network Indicators:
- Monitor for unusual traffic patterns or requests targeting TensorFlow services that could indicate exploitation attempts, though this is less specific.
SIEM Query:
Example query for logs: 'source="tensorflow.log" AND (error OR crash OR "null pointer" OR "RandomShuffle")'🔗 References
- https://github.com/tensorflow/tensorflow/commit/728113a3be690facad6ce436660a0bc1858017fa
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gf97-q72m-7579
- https://github.com/tensorflow/tensorflow/commit/728113a3be690facad6ce436660a0bc1858017fa
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gf97-q72m-7579
