CVE-2021-29587 : Vulnerability Insights and Analysis

TensorFlow is an end-to-end open source platform for machine learning. The

Prepare

step of the

SpaceToDepth

TFLite operator does not check for 0 before division. An attacker can craft a model to exploit this vulnerability. The fix is included in TensorFlow 2.5.0 and also backported to versions 2.4.2, 2.3.3, 2.2.3, and 2.1.4.

Understanding CVE-2021-29587

This CVE describes a vulnerability in TensorFlow that allows an attacker to trigger a division by zero in TFLite's implementation of the

SpaceToDepth

operator.

What is CVE-2021-29587?

The CVE-2021-29587 vulnerability in TensorFlow arises due to a missing check for zero before division in the

Prepare

step of the

SpaceToDepth

TFLite operator. This can be exploited by attackers to craft malicious models.

The Impact of CVE-2021-29587

With a CVSS base score of 2.5 (Low severity), the vulnerability has a high attack complexity and requires low privileges. The attack vector is local, affecting confidentiality and integrity minimally.

Technical Details of CVE-2021-29587

The vulnerability allows attackers to trigger a division by zero in TensorFlow's TFLite

SpaceToDepth

implementation, potentially leading to denial of service or other security compromises.

Vulnerability Description

Attackers can exploit the vulnerable

SpaceToDepth

operator by crafting a model to set

params->block_size

to zero, triggering a division by zero.

Affected Systems and Versions

The affected versions of TensorFlow include < 2.1.4, >= 2.2.0, < 2.2.3, >= 2.3.0, < 2.3.3, and >= 2.4.0, < 2.4.2.

Exploitation Mechanism

By manipulating the

params->block_size

value in a crafted model, an attacker can force the division by zero in the

SpaceToDepth

TFLite operator.

Mitigation and Prevention

To mitigate CVE-2021-29587:

Immediate Steps to Take

Update TensorFlow to versions where the fix is included (2.5.0 or backported versions).

Long-Term Security Practices

Regularly update TensorFlow to the latest versions and stay informed about security advisories.

Patching and Updates

Apply patches provided by TensorFlow to address the vulnerability and ensure the software is up-to-date.