CVE-2023-25659
📋 TL;DR
This vulnerability in TensorFlow allows an out-of-bounds read when the DynamicStitch operation receives mismatched indices and data shapes, potentially leading to memory corruption or information disclosure. It affects users of TensorFlow versions before 2.12.0 and 2.11.1 who run untrusted models or inputs.
💻 Affected Systems
- TensorFlow
📦 What is this software?
⚠️ Risk & Real-World Impact
Worst Case
Exploitation could cause a denial of service via application crash, memory corruption enabling arbitrary code execution, or leakage of sensitive data from memory.
Likely Case
Most probable impact is a denial of service through application crashes or instability, as exploitation requires specific input manipulation.
If Mitigated
With proper input validation and sandboxing, impact is limited to denial of service or minimal information disclosure.
🎯 Exploit Status
Exploitation requires crafting inputs to trigger the shape mismatch; no public exploits are known, but the vulnerability is documented in advisories.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: TensorFlow 2.12.0 or 2.11.1
Vendor Advisory: https://github.com/tensorflow/tensorflow/security/advisories/GHSA-93vr-9q9m-pj8p
Restart Required: Yes
Instructions:
1. Update TensorFlow to version 2.12.0 or 2.11.1 using pip: 'pip install --upgrade tensorflow==2.12.0' or 'pip install --upgrade tensorflow==2.11.1'. 2. Restart any applications or services using TensorFlow to apply the fix.
🔧 Temporary Workarounds
Input Validation for DynamicStitch
allImplement custom checks to ensure indices and data parameters match expected shapes before calling DynamicStitch.
# Example Python code to validate shapes
import tensorflow as tf
# Add validation logic before tf.raw_ops.DynamicStitch call
🧯 If You Can't Patch
- Restrict access to TensorFlow operations to trusted users only.
- Monitor and log usage of DynamicStitch for anomalous inputs and implement runtime sandboxing.
🔍 How to Verify
Check if Vulnerable:
Check TensorFlow version: if below 2.12.0 and not 2.11.1, it is vulnerable. Review code for DynamicStitch usage with untrusted inputs.Check Version:
python -c 'import tensorflow as tf; print(tf.__version__)'Verify Fix Applied:
After updating, confirm version is 2.12.0 or 2.11.1 and test DynamicStitch with various inputs to ensure no crashes or errors.📡 Detection & Monitoring
Log Indicators:
- Log entries indicating TensorFlow crashes, segmentation faults, or errors related to DynamicStitch operations.
Network Indicators:
- Unusual network traffic patterns to TensorFlow services with crafted payloads, though specific signatures are not defined.
SIEM Query:
Example: 'event_source:"TensorFlow" AND (error:"DynamicStitch" OR crash)'🔗 References
- https://github.com/tensorflow/tensorflow/commit/ee004b18b976eeb5a758020af8880236cd707d05
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-93vr-9q9m-pj8p
- https://github.com/tensorflow/tensorflow/commit/ee004b18b976eeb5a758020af8880236cd707d05
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-93vr-9q9m-pj8p
