CVE-2021-29584: TensorFlow integer overflow DoS

CISO Take

A local attacker with low privileges can crash TensorFlow processes by crafting a SparseSplit operation with oversized tensor dimensions, triggering an integer overflow and CHECK-fail assertion. In shared ML platforms—Jupyter hubs, multi-tenant model servers, or internal AI APIs accepting user-defined ops—this becomes a practical availability threat. Upgrade to TF 2.5.0 or patched backports (2.4.2, 2.3.3, 2.2.3, 2.1.4) immediately; restrict untrusted tensor op submission in shared environments as a compensating control.

What is the risk?

Medium overall, but context-dependent. The CVSS 5.5 score reflects local access requirement, which limits mass exploitation. However, in shared ML infrastructure (multi-tenant Jupyter, internal model serving endpoints, AutoML platforms), 'local access' is often a low bar—any authenticated user or API consumer qualifies. No in-the-wild exploitation recorded, not in CISA KEV. The trivial exploitation mechanics (craft oversized shape dimensions) elevate practical risk above the base score in shared-compute environments.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
195.8K OpenSSF 7.1 3.7K dependents Pushed 3d ago 4% patched ~1372d to patch Full package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

5.5 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 9% 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?

AV Local

AC Low

PR Low

UI None

S Unchanged

C None

I None

A High

What should I do?

5 steps

PATCH

Upgrade TensorFlow to ≥2.5.0 or apply backports: 2.4.2, 2.3.3, 2.2.3, or 2.1.4. Commit fix: 4c0ee937c0f61c4fc5f5d32d9bb4c67428012a60.
VALIDATE INPUT

Add dimension bounds checks before passing user-supplied shapes to SparseSplit; reject requests where shape dimensions exceed safe thresholds.
ISOLATE

Run TF model servers in separate processes per user/tenant to contain blast radius of a crash.
MONITOR

Alert on abnormal TF process terminations (SIGABRT/CHECK-fail logs) as potential exploitation indicators.
DETECT

Search for patterns like extreme tensor dimension values (approaching INT64_MAX) in model serving logs.

How is it classified?

DoS Framework Inference AML.T0010.001 - AI Software AML.T0029 - Denial of AI Service 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 for high-risk AI

ISO 42001

A.9.2 - AI System Availability and Resilience

NIST AI RMF

GOVERN-1.4 - Organizational teams are committed to risk management MANAGE-2.2 - Mechanisms to sustain AI risk management

Frequently Asked Questions

What is CVE-2021-29584?

A local attacker with low privileges can crash TensorFlow processes by crafting a SparseSplit operation with oversized tensor dimensions, triggering an integer overflow and CHECK-fail assertion. In shared ML platforms—Jupyter hubs, multi-tenant model servers, or internal AI APIs accepting user-defined ops—this becomes a practical availability threat. Upgrade to TF 2.5.0 or patched backports (2.4.2, 2.3.3, 2.2.3, 2.1.4) immediately; restrict untrusted tensor op submission in shared environments as a compensating control.

Is CVE-2021-29584 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2021-29584, increasing the risk of exploitation.

How to fix CVE-2021-29584?

1. PATCH: Upgrade TensorFlow to ≥2.5.0 or apply backports: 2.4.2, 2.3.3, 2.2.3, or 2.1.4. Commit fix: 4c0ee937c0f61c4fc5f5d32d9bb4c67428012a60. 2. VALIDATE INPUT: Add dimension bounds checks before passing user-supplied shapes to SparseSplit; reject requests where shape dimensions exceed safe thresholds. 3. ISOLATE: Run TF model servers in separate processes per user/tenant to contain blast radius of a crash. 4. MONITOR: Alert on abnormal TF process terminations (SIGABRT/CHECK-fail logs) as potential exploitation indicators. 5. DETECT: Search for patterns like extreme tensor dimension values (approaching INT64_MAX) in model serving logs.

What systems are affected by CVE-2021-29584?

This vulnerability affects the following AI/ML architecture patterns: training pipelines, model serving, shared ML platforms, notebook environments.

What is the CVSS score for CVE-2021-29584?

CVE-2021-29584 has a CVSS v3.1 base score of 5.5 (MEDIUM). The EPSS exploitation probability is 0.19%.

What is the AI security impact?

Affected AI Architectures

training pipelinesmodel servingshared ML platformsnotebook environments

MITRE ATLAS Techniques

AML.T0010.001 AI Software

AML.T0029 Denial of AI Service

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Art.15

ISO 42001: A.9.2

NIST AI RMF: GOVERN-1.4, MANAGE-2.2

What are the technical details?

Original Advisory

TensorFlow is an end-to-end open source platform for machine learning. An attacker can trigger a denial of service via a `CHECK`-fail in caused by an integer overflow in constructing a new tensor shape. This is because the implementation(https://github.com/tensorflow/tensorflow/blob/0908c2f2397c099338b901b067f6495a5b96760b/tensorflow/core/kernels/sparse_split_op.cc#L66-L70) builds a dense shape without checking that the dimensions would not result in overflow. The `TensorShape` constructor(https://github.com/tensorflow/tensorflow/blob/6f9896890c4c703ae0a0845394086e2e1e523299/tensorflow/core/framework/tensor_shape.cc#L183-L188) uses a `CHECK` operation which triggers when `InitDims`(https://github.com/tensorflow/tensorflow/blob/6f9896890c4c703ae0a0845394086e2e1e523299/tensorflow/core/framework/tensor_shape.cc#L212-L296) returns a non-OK status. This is a legacy implementation of the constructor and operations should use `BuildTensorShapeBase` or `AddDimWithStatus` to prevent `CHECK`-failures in the presence of overflows. 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 internal data scientist on a shared ML platform submits a SparseSplit operation via a Jupyter notebook with a dense_shape tensor constructed so that multiplying its dimensions causes integer overflow (e.g., shape [9223372036854775807, 2]). TensorShape constructor calls InitDims, which returns a non-OK status due to overflow; the CHECK macro triggers, sending SIGABRT to the TF process. In a shared model-serving scenario, a single crafted API request to an endpoint accepting sparse tensor inputs crashes the serving process, causing a denial of service for all concurrent users. Requires no exploit payload, just knowledge of the overflow condition.

Weaknesses (CWE)

CWE-190 Integer Overflow or Wraparound

CWE-190 — Integer Overflow or Wraparound: The product performs a calculation that can produce an integer overflow or wraparound when the logic assumes that the resulting value will always be larger than the original value. This occurs when an integer value is incremented to a value that is too large to store in the associated representation. When this occurs, the value may become a very small or negative number.

[Requirements] Ensure that all protocols are strictly defined, such that all out-of-bounds behavior can be identified simply, and require strict conformance to the protocol.
[Requirements] Use a language that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid. If possible, choose a language or compiler that performs automatic bounds checking.

Source: MITRE CWE corpus.