AVID-2026-R0912

Description

Access to invalid memory during shape inference in Cudnn* ops (CVE-2021-41221)

Details

TensorFlow is an open source platform for machine learning. In affected versions the shape inference code for the Cudnn* operations in TensorFlow can be tricked into accessing invalid memory, via a heap buffer overflow. This occurs because the ranks of the input, input_h and input_c parameters are not validated, but code assumes they have certain values. The fix will be included in TensorFlow 2.7.0. We will also cherrypick this commit on TensorFlow 2.6.1, TensorFlow 2.5.2, and TensorFlow 2.4.4, as these are also affected and still in supported range.

Reason for inclusion in AVID: The CVE describes a software vulnerability in TensorFlow’s AI software stack (Cudnn* shape inference) that can cause invalid memory access via a heap buffer overflow. It affects a component (TensorFlow) used to build, train, and deploy AI systems, making it relevant to the general-purpose AI supply chain and security posture.

References

Affected or Relevant Artifacts

Developer: tensorflow
Deployer: tensorflow
Artifact Details:

Type	Name
System	tensorflow

Impact

AVID Taxonomy Categorization

Risk domains: Security
SEP subcategories: S0100: Software Vulnerability
Lifecycle stages: L06: Deployment

CVSS

Version	3.1
Vector String	CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
Base Score	7.8
Base Severity	🔴 High
Attack Vector	LOCAL
Attack Complexity	🟢 Low
Privileges Required	🟢 Low
User Interaction	NONE
Scope	UNCHANGED
Confidentiality Impact	🔴 High
Integrity Impact	🔴 High
Availability Impact	🔴 High

CWE

ID	Description
CWE-120	CWE-120: Buffer Copy without Checking Size of Input (‘Classic Buffer Overflow’)

Other information

Report Type: Advisory
Credits:
Date Reported: 2021-11-05
Version: 0.3.3
AVID Entry