Critical AnythingLLM Vulnerability Exposes Systems to Remote Code Execution

Critical AnythingLLM Vulnerability Exposes Systems to Remote Code Execution

A critical security flaw (CVE-2024-13059) in the open-source AI framework AnythingLLM has raised alarms across cybersecurity communities.

The vulnerability, discovered in February 2025, allows attackers with administrative privileges to execute malicious code remotely, potentially compromising entire systems.

DetailDescription
CVE IDCVE-2024-13059
SeverityCritical (CVSS 9.1)
EPSS Score0.04% (Low exploitation probability)
Affected VersionsAnythingLLM versions < 1.3.1
Patched Version1.3.1 (released February 10, 2025)
ImpactRemote Code Execution (RCE) via path traversal

Vulnerability Breakdown

According to the Offsec report, the flaw originates from improper filename sanitization in the multer library, used by AnythingLLM for file uploads.

– Advertisement –

Non-ASCII filenames containing directory traversal sequences (e.g., ../../malicious.sh) are not properly validated, enabling attackers to write files to unintended locations.

The exploitation process begins when an attacker with manager or admin access to a vulnerable AnythingLLM instance crafts a malicious file containing directory traversal sequences in its filename (e.g., ../../malicious.js).

Due to improper sanitization of non-ASCII filenames in the multer library, the application fails to detect these sequences.

When the file is uploaded via AnythingLLM’s interface, the server mistakenly writes it to unintended locations outside the restricted upload directory.

For example, a file named ../../../etc/cron.d/exploit could be placed in a system directory responsible for scheduling automated tasks.

If the file is executed—either through scheduled jobs, startup scripts, or other mechanisms—it enables full remote code execution (RCE), granting the attacker control over the compromised system.

This attack vector highlights how elevated privileges and insufficient input validation combine to turn a simple file upload into a critical system breach.

Organizations using AnythingLLM for AI-driven workflows, such as customer support automation or internal data analysis—are urged to act swiftly.

Delayed patching could lead to data breaches, service disruptions, or unauthorized access to sensitive infrastructure.

Find this News Interesting! Follow us on Google News, LinkedIn, & X to Get Instant Updates!


Source link