CVE-2021-29513: TensorFlow: type confusion → null ptr deref (CVSS 7.8)

HIGHPoC AVAILABLE

Published May 14, 2021

CISO Take

Any TensorFlow deployment that accepts external tensor inputs is at risk of process crash or potential code execution via type confusion in the Python-to-C++ conversion layer. Patch to TF 2.5.0 (or 2.4.2/2.3.3/2.2.3/2.1.4 for pinned versions) immediately. Multi-tenant ML platforms and model serving APIs with untrusted input are highest priority.

What is the risk?

CVSS 7.8 High with local attack vector and low privilege requirement. The full CIA impact score (C:H/I:H/A:H) suggests exploitation potential beyond simple DoS — type confusion vulnerabilities can enable memory corruption and code execution. Not in CISA KEV with no confirmed active exploitation, but trivial exploitation complexity elevates practical risk in shared ML environments significantly.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
201.3KOpenSSF 7.63.4K dependentsPushed 5d ago4% patched~1372d to patchFull package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

7.8 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 10% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVLocal

ACLow

PRLow

UINone

SUnchanged

CHigh

IHigh

AHigh

What should I do?

5 steps

Upgrade TensorFlow to 2.5.0, or cherry-picked branches 2.4.2/2.3.3/2.2.3/2.1.4.
Enforce numeric dtype validation before passing tensors to TF operations at all API boundaries.
Run TF inference processes in isolated containers to limit blast radius.
Monitor for unexpected TF process crashes in serving infrastructure.
Audit any externally-accessible TensorFlow Serving endpoints for unauthenticated tensor submission.

How is it classified?

Code ExecutionDoSFrameworkInferenceAML.T0029- Denial of AI Service AML.T0043- Craft Adversarial Data AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15-Accuracy, robustness and cybersecurity

ISO 42001

8.4-AI system risk treatment

NIST AI RMF

MANAGE-2.2-Mechanisms to maintain AI system operations

OWASP LLM Top 10

LLM05:2025-Improper Output Handling

Frequently Asked Questions

What is CVE-2021-29513?

Is CVE-2021-29513 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2021-29513, increasing the risk of exploitation.

How to fix CVE-2021-29513?

1. Upgrade TensorFlow to 2.5.0, or cherry-picked branches 2.4.2/2.3.3/2.2.3/2.1.4. 2. Enforce numeric dtype validation before passing tensors to TF operations at all API boundaries. 3. Run TF inference processes in isolated containers to limit blast radius. 4. Monitor for unexpected TF process crashes in serving infrastructure. 5. Audit any externally-accessible TensorFlow Serving endpoints for unauthenticated tensor submission.

What systems are affected by CVE-2021-29513?

This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, inference, multi-tenant ML platforms.

What is the CVSS score for CVE-2021-29513?

CVE-2021-29513 has a CVSS v3.1 base score of 7.8 (HIGH). The EPSS exploitation probability is 0.20%.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesinferencemulti-tenant ML platforms

MITRE ATLAS Techniques

AML.T0029Denial of AI Service

AML.T0043Craft Adversarial Data

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: 8.4

NIST AI RMF: MANAGE-2.2

OWASP LLM Top 10: LLM05:2025

What are the technical details?

Original Advisory

TensorFlow is an end-to-end open source platform for machine learning. Calling TF operations with tensors of non-numeric types when the operations expect numeric tensors result in null pointer dereferences. The conversion from Python array to C++ array(https://github.com/tensorflow/tensorflow/blob/ff70c47a396ef1e3cb73c90513da4f5cb71bebba/tensorflow/python/lib/core/ndarray_tensor.cc#L113-L169) is vulnerable to a type confusion. The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, as these are also affected and still in supported range.

Exploitation Scenario

An attacker with low-privilege access to a shared ML platform (Jupyter user, API client) submits a tensor with string or boolean dtype to a TF operation expecting float/int. The ndarray_tensor.cc Python-to-C++ conversion path dereferences a null pointer, crashing the TF process. In a worst-case memory corruption scenario, controlled exploitation achieves code execution within the ML serving process, enabling pivot to model weights, training data, or adjacent services on the host.

Weaknesses (CWE)

CWE-476NULL Pointer DereferencePrimary CWE-843Access of Resource Using Incompatible TypePrimary CWE-476NULL Pointer Dereference

CWE-476 — NULL Pointer Dereference: The product dereferences a pointer that it expects to be valid but is NULL.

[Implementation] For any pointers that could have been modified or provided from a function that can return NULL, check the pointer for NULL before use. When working with a multithreaded or otherwise asynchronous environment, ensure that proper locking APIs are used to lock before the check, and unlock when it has finished [REF-1484].
[Requirements] Select a programming language that is not susceptible to these issues.

Source: MITRE CWE corpus.