CVE-2021-29552: 
Python vulnerability analysis and mitigation

TensorFlow, an end-to-end open source platform for machine learning, was found to contain a vulnerability (CVE-2021-29552) related to the UnsortedSegmentJoin operation. The vulnerability was discovered by Ying Wang and Yakun Zhang of Baidu X-Team and affects TensorFlow versions prior to 2.5.0, including versions 2.1.x, 2.2.x, 2.3.x, and 2.4.x (GitHub Advisory).

The vulnerability stems from an implementation assumption in the UnsortedSegmentJoin operation where the numsegments tensor is expected to be a valid scalar. When the tensor is empty, the CHECK validation in .scalar()() fails as it requires exactly one element, resulting in process termination. The issue is present in the implementation at tensorflow/core/kernels/unsortedsegmentjoinop.cc. The vulnerability has been assigned a CVSS v3.1 base score of 5.5 (MEDIUM) by NVD with vector CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H (NVD).

An attacker can trigger a denial of service condition by manipulating the values of the num_segments tensor argument for UnsortedSegmentJoin. The exploitation leads to process termination, affecting the availability of the TensorFlow application (GitHub Advisory).

The vulnerability has been patched in TensorFlow 2.5.0. The fix has also been backported to versions 2.4.2, 2.3.3, 2.2.3, and 2.1.4. Users are advised to upgrade to these patched versions. The fix implements additional validation to check that the number of segments cannot be empty (GitHub Commit).