CVE-2019-9635: TensorFlow NULL ptr deref DoS

CISO Take

This 2019 vulnerability allows an attacker to crash TensorFlow processes by submitting a malformed GIF file, resulting in denial of service to any image-processing ML pipeline. Any TensorFlow deployment below 1.12.2 handling image inputs should be patched immediately — though in 2026 this should already be resolved in any maintained environment. Verify your TensorFlow versions across inference infrastructure and ensure input validation exists at API boundaries.

What is the risk?

Low-to-medium risk in current environments. The vulnerability is limited to availability impact (DoS) with no code execution or data exfiltration component. Exploitability is trivial — a single malformed GIF triggers the crash. Primary concern is in production inference APIs accepting unvalidated image uploads; a crash loop could degrade ML service availability. Any TensorFlow version >= 1.12.2 is not affected. Given the age (2019) and public patch availability, residual risk exists only in legacy or unpatched deployments.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
TensorFlow	pip	—	No patch
195.8K OpenSSF 7.1 3.7K dependents Pushed 2d ago 4% patched ~1372d to patch Full package profile →

Do you use TensorFlow? You're affected.

How severe is it?

CVSS 3.1

N/A

EPSS

0.5%

chance of exploitation in 30 days

Higher than 36% 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 should I do?

6 steps

Patch: Upgrade TensorFlow to 1.12.2 or later immediately.
Input validation: Implement server-side validation of uploaded files — verify magic bytes, reject malformed images before passing to TensorFlow.
Process isolation: Run inference workers in isolated containers/processes with automatic restart policies to minimize DoS window.
Rate limiting: Apply rate limits on image upload endpoints to reduce crash-loop exploitation.
Detection: Monitor for abnormal TensorFlow process termination events and correlate with incoming request payloads.
Inventory: Audit all TensorFlow versions across inference servers, training infrastructure, and CI/CD pipelines.

How is it classified?

DoS Framework Inference AML.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 operation and monitoring

NIST AI RMF

RES-1.1 - Resilience — AI system reliability under adverse conditions

OWASP LLM Top 10

LLM10 - Model Denial of Service

Frequently Asked Questions

What is CVE-2019-9635?

This 2019 vulnerability allows an attacker to crash TensorFlow processes by submitting a malformed GIF file, resulting in denial of service to any image-processing ML pipeline. Any TensorFlow deployment below 1.12.2 handling image inputs should be patched immediately — though in 2026 this should already be resolved in any maintained environment. Verify your TensorFlow versions across inference infrastructure and ensure input validation exists at API boundaries.

Is CVE-2019-9635 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2019-9635, increasing the risk of exploitation.

How to fix CVE-2019-9635?

1. Patch: Upgrade TensorFlow to 1.12.2 or later immediately. 2. Input validation: Implement server-side validation of uploaded files — verify magic bytes, reject malformed images before passing to TensorFlow. 3. Process isolation: Run inference workers in isolated containers/processes with automatic restart policies to minimize DoS window. 4. Rate limiting: Apply rate limits on image upload endpoints to reduce crash-loop exploitation. 5. Detection: Monitor for abnormal TensorFlow process termination events and correlate with incoming request payloads. 6. Inventory: Audit all TensorFlow versions across inference servers, training infrastructure, and CI/CD pipelines.

What systems are affected by CVE-2019-9635?

This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, inference endpoints.

What is the CVSS score for CVE-2019-9635?

No CVSS score has been assigned yet.

What is the AI security impact?

Affected AI Architectures

model servingtraining pipelinesinference endpoints

MITRE ATLAS Techniques

AML.T0029 Denial of AI Service

AML.T0043 Craft Adversarial Data

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: 8.4

NIST AI RMF: RES-1.1

OWASP LLM Top 10: LLM10

What are the technical details?

Original Advisory

NULL pointer dereference in Google TensorFlow before 1.12.2 could cause a denial of service via an invalid GIF file.

Exploitation Scenario

An adversary targeting an organization's image classification API (e.g., a content moderation or medical imaging service powered by TensorFlow) crafts or obtains a malformed GIF file that triggers the NULL pointer dereference. They submit this file via the public-facing upload endpoint. The TensorFlow process crashes, taking down the inference service. If the service lacks automatic restart or circuit-breaking logic, this results in sustained unavailability. The attacker can automate repeated submissions to maintain the DoS state, disrupting business operations dependent on the ML service.

Weaknesses (CWE)

CWE-476 NULL Pointer Dereference Primary

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.