CVE-2023-25664: TensorFlow has Heap-buffer-overflow in AvgPoolGrad
7.5
CVSS Score
3.1
Basic Information
CVE ID
GHSA ID
EPSS Score
0.21645%
CWE
Published
3/24/2023
Updated
3/27/2023
KEV Status
No
Technology
Python
Technical Details
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
| Package Name | Ecosystem | Vulnerable Versions | First Patched Version |
|---|---|---|---|
| tensorflow | pip | < 2.11.1 | 2.11.1 |
| tensorflow-cpu | pip | < 2.11.1 | 2.11.1 |
Vulnerability Intelligence
Miggo AI
Root Cause Analysis
The vulnerability stems from missing input validation in the AvgPoolGrad implementation. The commit ddaac2b added validation loops using GetBroadcastSize to check kernel/stride compatibility before buffer operations. The original code proceeded to memory operations without these checks, allowing invalid parameter combinations to cause buffer overflows. The test case added in pooling_ops_test.py specifically targets stride validation, and the CWE-122 (heap overflow) classification confirms memory safety issues in gradient calculation.