Learn about CVE-2021-41198 affecting TensorFlow versions 2.4.4, 2.5.0-2.5.2, and 2.6.0-2.6.1. Discover the impact, mitigation steps, and how to prevent the overflow vulnerability.
TensorFlow is an open-source platform for machine learning. In affected versions, calling
tf.tile
with a large input argument results in a process crash due to an overflow. This issue affects versions >= 2.6.0, < 2.6.1; >= 2.5.0, < 2.5.2; and < 2.4.4. The impact includes a high availability impact with a CVSS base score of 5.5.
Understanding CVE-2021-41198
In this CVE, the
tf.tile
function in TensorFlow triggers a process crash when tiled with a large tensor due to an overflow condition.
What is CVE-2021-41198?
tf.tile
is used with a large input, causing an overflow due to an element count exceeding int64_t
range.The Impact of CVE-2021-41198
Technical Details of CVE-2021-41198
This section covers the technical aspects of the vulnerability.
Vulnerability Description
tf.tile
is used with a large input, leading to a process crash from an overflow detected by a CHECK
statement.Affected Systems and Versions
Exploitation Mechanism
tf.tile
excessively with a significant input tensor triggers the overflow, detected by a CHECK
statement, resulting in a process crash.Mitigation and Prevention
Suggestions to mitigate the impact and prevent future occurrences.
Immediate Steps to Take
Long-Term Security Practices
Patching and Updates