Discover the details of CVE-2022-23591, a critical vulnerability in TensorFlow leading to a stack overflow issue. Learn about impact, affected versions, and mitigation steps.
TensorFlow is an open-source machine learning framework that has been affected by a critical vulnerability leading to a stack overflow. Learn more about the impact, technical details, and mitigation steps related to CVE-2022-23591.
Understanding CVE-2022-23591
This section provides insights into the nature of the vulnerability and its potential impact.
What is CVE-2022-23591?
TensorFlow's
GraphDef
format restriction violation causes a stack overflow issue during execution when loading a SavedModel
, affecting versions 2.5.3 to 2.7.0.
The Impact of CVE-2022-23591
The vulnerability poses a high availability impact with a CVSS base score of 7.5, categorized as HIGH severity due to uncontrolled resource consumption.
Technical Details of CVE-2022-23591
Explore the specific technical aspects of the vulnerability in this section.
Vulnerability Description
The vulnerability arises from consuming a
GraphDef
containing self-recursive functions, leading to a stack overflow at execution time.
Affected Systems and Versions
TensorFlow versions >= 2.7.0, < 2.7.1, >= 2.6.0, < 2.6.3, and < 2.5.3 are impacted by this vulnerability.
Exploitation Mechanism
The vulnerability can be exploited when loading a
SavedModel
that contains a GraphDef
with specific self-recursive fragments.
Mitigation and Prevention
Discover the necessary steps to address and prevent exploitation of CVE-2022-23591.
Immediate Steps to Take
Users are advised to update TensorFlow to version 2.8.0 to mitigate the vulnerability. For versions 2.5.3 to 2.7.0, patches are available in TensorFlow 2.7.1, TensorFlow 2.6.3, and TensorFlow 2.5.3.
Long-Term Security Practices
Regularly update TensorFlow to the latest versions and follow security best practices to minimize the risk of similar vulnerabilities.
Patching and Updates
Stay informed about security advisories and apply patches promptly to ensure a secure TensorFlow environment.