CVE-2021-37668 : Security Advisory and Response

TensorFlow is an end-to-end open source platform for machine learning. In affected versions, an attacker can cause denial of service by triggering a division by 0 in applications serving models using

tf.raw_ops.UnravelIndex

.

Understanding CVE-2021-37668

This CVE involves a vulnerability in TensorFlow Lite's

tf.raw_ops.UnravelIndex

that allows an attacker to trigger a division by 0, leading to a denial of service.

What is CVE-2021-37668?

In TensorFlow versions >= 2.3.4 and < 2.5.1, a flaw in the implementation of

tf.raw_ops.UnravelIndex

can be exploited to cause a division by 0, resulting in denial of service.

The Impact of CVE-2021-37668

The impact of this vulnerability is rated as MEDIUM with a CVSS base score of 5.5. Attack complexity is LOW, with LOCAL attack vector and HIGH availability impact.

Technical Details of CVE-2021-37668

The vulnerability arises due to the lack of checking whether the tensor subsumed by

dims

is not empty. This flaw allows an attacker to trigger a division by 0.

Vulnerability Description

The flaw in TensorFlow allows an attacker to exploit

tf.raw_ops.UnravelIndex

to divide by 0 and cause a denial of service.

Affected Systems and Versions

Affected versions include TensorFlow >= 2.5.0, < 2.5.1, >= 2.4.0, < 2.4.3, and < 2.3.4.

Exploitation Mechanism

By triggering a division by 0 in the implementation of

tf.raw_ops.UnravelIndex

, an attacker can exploit this vulnerability.

Mitigation and Prevention

It is crucial to take immediate steps, implement long-term security practices, and apply available patches and updates.

Immediate Steps to Take

Users are advised to update to TensorFlow 2.6.0 to mitigate this vulnerability and to apply the provided patches to TensorFlow 2.5.1, 2.4.3, and 2.3.4.

Long-Term Security Practices

Maintain updated software, conduct regular security audits, and adhere to best practices to enhance system security.

Patching and Updates

Ensure timely application of security patches provided by TensorFlow to address this vulnerability.