CVE-2022-36019 — HIGH (CVSS 7.5) AI Security Vulnerability

CISO Take

A network-accessible denial-of-service vulnerability in TensorFlow's per-channel quantization operator allows any unauthenticated attacker to crash model serving infrastructure by sending a malformed inference request. If your organization exposes TensorFlow Serving endpoints or uses TF 2.7.x–2.9.x in production pipelines, patching to TF 2.10.0 (or the respective backport) is the immediate action. No authentication is required to exploit this, making exposed serving endpoints a direct target.

Risk Assessment

High risk for organizations with externally or internally exposed TensorFlow Serving endpoints. CVSS 7.5 with AV:N/AC:L/PR:N/UI:N means zero-barrier exploitation over the network. The attack requires no ML knowledge—only the ability to craft a tensor with incorrect rank. The availability-only impact (C:N/I:N/A:H) limits blast radius to service disruption rather than data breach, but in SLA-bound production ML environments this translates directly to revenue loss and SLA violations. No active exploitation in the wild observed; not in CISA KEV.

Affected Systems

Package	Ecosystem	Vulnerable Range	Patched
tensorflow	pip	—	No patch
195.0K OpenSSF 7.2 3.7K dependents Pushed 6d ago 4% patched ~1372d to patch Full package profile →

Do you use tensorflow? You're affected.

Severity & Risk

CVSS 3.1

7.5 / 10

EPSS

0.1%

chance of exploitation in 30 days

Higher than 20% 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, CISA SSVC, EPSS, trickest/cve, and Nuclei templates.

Attack Surface

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

Recommended Action

5 steps

PATCH

Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 immediately. Commit 785d67a78a1d533759fcd2f5e8d6ef778de849e0 is the authoritative fix.
INPUT VALIDATION

Enforce tensor rank validation at API boundaries before ops execute—reject any request where min/max tensors for FakeQuantWithMinMaxVarsPerChannel are not rank-1.
NETWORK CONTROLS

Restrict TensorFlow Serving exposure behind authenticated proxies; avoid direct internet exposure of gRPC/REST serving ports.
MONITORING

Alert on abnormal process restarts of TF serving workers.
DETECTION

Log and rate-limit inference requests with unusual tensor shapes.

CISA SSVC Assessment

Decision Track

Exploitation none

Automatable No

Technical Impact partial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

Classification

DoS Framework Inference AML.T0029 - Denial of AI Service AML.T0043 - Craft Adversarial Data AML.T0049 - Exploit Public-Facing Application

Compliance Impact

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6 - AI system security A.9.3 - Reliability and availability of AI systems

NIST AI RMF

MANAGE-2.2 - Mechanisms are in place to inventory AI risks

OWASP LLM Top 10

LLM04 - Model Denial of Service

Frequently Asked Questions

What is CVE-2022-36019?

A network-accessible denial-of-service vulnerability in TensorFlow's per-channel quantization operator allows any unauthenticated attacker to crash model serving infrastructure by sending a malformed inference request. If your organization exposes TensorFlow Serving endpoints or uses TF 2.7.x–2.9.x in production pipelines, patching to TF 2.10.0 (or the respective backport) is the immediate action. No authentication is required to exploit this, making exposed serving endpoints a direct target.

Is CVE-2022-36019 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2022-36019, increasing the risk of exploitation.

How to fix CVE-2022-36019?

1. PATCH: Upgrade to TensorFlow 2.10.0, 2.9.1, 2.8.1, or 2.7.2 immediately. Commit 785d67a78a1d533759fcd2f5e8d6ef778de849e0 is the authoritative fix. 2. INPUT VALIDATION: Enforce tensor rank validation at API boundaries before ops execute—reject any request where min/max tensors for FakeQuantWithMinMaxVarsPerChannel are not rank-1. 3. NETWORK CONTROLS: Restrict TensorFlow Serving exposure behind authenticated proxies; avoid direct internet exposure of gRPC/REST serving ports. 4. MONITORING: Alert on abnormal process restarts of TF serving workers. 5. DETECTION: Log and rate-limit inference requests with unusual tensor shapes.

What systems are affected by CVE-2022-36019?

This vulnerability affects the following AI/ML architecture patterns: model serving, training pipelines, quantization-aware training, edge model deployment pipelines, inference endpoints.

What is the CVSS score for CVE-2022-36019?

CVE-2022-36019 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.06%.

Technical Details

NVD Description

TensorFlow is an open source platform for machine learning. If `FakeQuantWithMinMaxVarsPerChannel` is given `min` or `max` tensors of a rank other than one, it results in a `CHECK` fail that can be used to trigger a denial of service attack. We have patched the issue in GitHub commit 785d67a78a1d533759fcd2f5e8d6ef778de849e0. The fix will be included in TensorFlow 2.10.0. We will also cherrypick this commit on TensorFlow 2.9.1, TensorFlow 2.8.1, and TensorFlow 2.7.2, as these are also affected and still in supported range. There are no known workarounds for this issue.

Exploitation Scenario

An adversary targeting an organization's model inference API identifies a TensorFlow Serving endpoint (default ports 8500/8501). They craft a minimal inference request that includes FakeQuantWithMinMaxVarsPerChannel inputs where the min or max tensor is rank-0 (scalar) or rank-2+ instead of the expected rank-1 vector. The malformed request triggers a CHECK assertion failure in TensorFlow's C++ runtime, immediately terminating the serving process. With a simple loop, the attacker can repeatedly crash the service faster than auto-restart mechanisms recover, achieving persistent denial of service against the production ML API with no credentials or ML expertise required.