CVE-2024-9056 — HIGH (CVSS 7.5) AI Security Vulnerability

Q: Is CVE-2024-9056 actively exploited?

No confirmed active exploitation of CVE-2024-9056 has been reported, but organizations should still patch proactively.

Q: How to fix CVE-2024-9056?

1. Inventory all BentoML deployments: 'pip show bentoml' on all serving hosts. 2. Place a reverse proxy (nginx/Caddy) or WAF in front of BentoML endpoints with strict multipart boundary length limits. 3. Enforce rate limiting and request size caps (e.g., client_max_body_size in nginx, body size limits in API gateway). 4. Restrict inference endpoint access via IP allowlisting where feasible. 5. Monitor GitHub advisory GHSA-hw8j-hw49-752c for upstream patch release. 6. Alert on anomalous CPU/memory spikes on BentoML inference servers as an indicator of active exploitation.

Q: What systems are affected by CVE-2024-9056?

This vulnerability affects the following AI/ML architecture patterns: model serving, inference APIs, MLOps pipelines, AI microservices.

Q: What is the CVSS score for CVE-2024-9056?

CVE-2024-9056 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.30%.

CISO Take

BentoML model serving endpoints are vulnerable to unauthenticated DoS via crafted HTTP multipart requests — no patch exists for versions <= 1.4.5. Any internet-exposed BentoML deployment risks complete inference service unavailability from a single attacker with no credentials. Immediately place a WAF or reverse proxy with multipart boundary length limits and rate limiting in front of all BentoML endpoints until upstream patches.

Risk Assessment

High severity (CVSS 7.5) with low EPSS (0.00151), indicating limited active exploitation evidence. However, the zero-authentication, zero-interaction attack vector makes this trivially weaponizable once a target is identified. Absence of an available patch elevates operational risk. Organizations exposing BentoML inference APIs directly to the internet or running multi-tenant AI serving platforms face the highest exposure surface.

Affected Systems

Package	Ecosystem	Vulnerable Range	Patched
bentoml	pip	<= 1.4.5	No patch
8.6K OpenSSF 6.3 22 dependents Pushed 9d ago 50% patched ~14d to patch Full package profile →

Do you use bentoml? You're affected.

Severity & Risk

CVSS 3.1

7.5 / 10

EPSS

0.3%

chance of exploitation in 30 days

Higher than 54% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○ CISA SSVC: Public PoC

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

6 steps

Inventory all BentoML deployments: 'pip show bentoml' on all serving hosts.
Place a reverse proxy (nginx/Caddy) or WAF in front of BentoML endpoints with strict multipart boundary length limits.
Enforce rate limiting and request size caps (e.g., client_max_body_size in nginx, body size limits in API gateway).
Restrict inference endpoint access via IP allowlisting where feasible.
Monitor GitHub advisory GHSA-hw8j-hw49-752c for upstream patch release.
Alert on anomalous CPU/memory spikes on BentoML inference servers as an indicator of active exploitation.

CISA SSVC Assessment

Decision Track*

Exploitation poc

Automatable Yes

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.T0034 - Cost Harvesting AML.T0049 - Exploit Public-Facing Application

Compliance Impact

This CVE is relevant to:

EU AI Act

Art. 15 - Accuracy, Robustness and Cybersecurity

ISO 42001

A.10.3 - AI System Operation and Monitoring

NIST AI RMF

MANAGE-2.4 - AI System Resilience and Incident Response

OWASP LLM Top 10

LLM10 - Unbounded Consumption

Frequently Asked Questions

What is CVE-2024-9056?

BentoML model serving endpoints are vulnerable to unauthenticated DoS via crafted HTTP multipart requests — no patch exists for versions <= 1.4.5. Any internet-exposed BentoML deployment risks complete inference service unavailability from a single attacker with no credentials. Immediately place a WAF or reverse proxy with multipart boundary length limits and rate limiting in front of all BentoML endpoints until upstream patches.

Is CVE-2024-9056 actively exploited?

No confirmed active exploitation of CVE-2024-9056 has been reported, but organizations should still patch proactively.

How to fix CVE-2024-9056?

1. Inventory all BentoML deployments: 'pip show bentoml' on all serving hosts. 2. Place a reverse proxy (nginx/Caddy) or WAF in front of BentoML endpoints with strict multipart boundary length limits. 3. Enforce rate limiting and request size caps (e.g., client_max_body_size in nginx, body size limits in API gateway). 4. Restrict inference endpoint access via IP allowlisting where feasible. 5. Monitor GitHub advisory GHSA-hw8j-hw49-752c for upstream patch release. 6. Alert on anomalous CPU/memory spikes on BentoML inference servers as an indicator of active exploitation.

What systems are affected by CVE-2024-9056?

This vulnerability affects the following AI/ML architecture patterns: model serving, inference APIs, MLOps pipelines, AI microservices.

What is the CVSS score for CVE-2024-9056?

CVE-2024-9056 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.30%.

Technical Details

NVD Description

BentoML version v1.3.4post1 is vulnerable to a Denial of Service (DoS) attack. The vulnerability can be exploited by appending characters, such as dashes (-), to the end of a multipart boundary in an HTTP request. This causes the server to continuously process each character, leading to excessive resource consumption and rendering the service unavailable. The issue is unauthenticated and does not require any user interaction, impacting all users of the service.

Exploitation Scenario

An adversary enumerates internet-facing BentoML inference APIs (e.g., via Shodan or targeted recon of an organization's AI product). Without credentials or prior access, they craft HTTP multipart POST requests with malformed boundary strings — appending hundreds or thousands of dashes to the boundary value. BentoML's file I/O descriptor processes each character sequentially, consuming CPU in proportion to boundary length. By sending concurrent malformed requests, the attacker exhausts server resources and renders the ML inference service unresponsive, effectively disabling any AI-powered application features relying on it.