CVE-2026-0558: Lollms Lollms DoS Vulnerability

CVE-2026-0558 Overview

A critical authentication bypass vulnerability has been discovered in parisneo/lollms, an AI-powered chatbot framework. The vulnerability exists in the /api/files/extract-text endpoint, which fails to enforce authentication requirements, allowing unauthenticated remote users to upload and process files without proper authorization. Unlike other file-related endpoints in the application, this specific endpoint lacks the Depends(get_current_active_user) dependency that should be enforcing access controls.

Critical Impact

Unauthenticated attackers can exploit this vulnerability to cause denial of service through resource exhaustion, potentially disclose sensitive information from processed files, and bypass the application's documented security policies.

Affected Products

• parisneo/lollms versions up to and including 2.2.0
• lollms AI chatbot framework (all versions prior to patch)
• Deployments using the vulnerable /api/files/extract-text endpoint

Discovery Timeline

• 2026-03-29 - CVE CVE-2026-0558 published to NVD
• 2026-03-31 - Last updated in NVD database

Technical Details for CVE-2026-0558

Vulnerability Analysis

This vulnerability represents a broken access control flaw (CWE-287: Improper Authentication) in the lollms file processing API. The /api/files/extract-text endpoint was implemented without the authentication middleware that protects other file-related operations in the application. This inconsistency in security controls creates an attack surface that allows unauthenticated users to interact with the file processing functionality.

The vulnerability can be exploited remotely over the network without any user interaction or prior authentication. Successful exploitation allows attackers to consume server resources by uploading and processing arbitrary files, potentially leading to service degradation or complete denial of service. Additionally, depending on the file types processed and server configuration, information disclosure may occur through the text extraction functionality.

Root Cause

The root cause of this vulnerability is the missing authentication dependency on the /api/files/extract-text endpoint in the backend/routers/files.py file. While other endpoints in the same file properly implement the Depends(get_current_active_user) function to verify that requests originate from authenticated users, this particular endpoint was inadvertently deployed without this security control. This oversight violates the application's documented security architecture and creates an exploitable authentication bypass.

Attack Vector

An attacker can exploit this vulnerability by sending direct HTTP requests to the /api/files/extract-text endpoint without providing any authentication credentials. The attack can be performed remotely from any network location that can reach the lollms instance. The attacker may:

1. Send repeated file upload requests to exhaust server CPU, memory, or disk resources
2. Upload specially crafted files designed to maximize processing time
3. Potentially extract sensitive information if the server processes files containing confidential data
4. Bypass rate limiting or other security controls that depend on user authentication

The following patch shows the security fix that adds authentication dependencies to the file endpoints:

 from docx.oxml import parse_xml
 from docx.oxml.ns import qn
 from latex2mathml.converter import convert as latex2mathml
-
+from PIL import Image
 
 # Try to import optional document parsing libraries
 try:

Source: GitHub Commit

Note: This is a partial patch snippet. The complete fix adds the Depends(get_current_active_user) dependency to enforce authentication on the vulnerable endpoint.

Detection Methods for CVE-2026-0558

Indicators of Compromise

• Unusual volume of requests to /api/files/extract-text endpoint from unauthenticated sources
• Spike in file processing activities without corresponding authenticated user sessions
• Server resource exhaustion (CPU, memory, disk) without legitimate user activity
• Log entries showing successful file uploads without associated authentication tokens

Detection Strategies

• Monitor HTTP access logs for requests to /api/files/extract-text that lack authentication headers or session cookies
• Implement anomaly detection for file upload patterns that deviate from normal authenticated user behavior
• Review application logs for text extraction operations occurring outside of authenticated sessions
• Deploy web application firewall rules to flag high-frequency requests to the vulnerable endpoint

Monitoring Recommendations

• Configure alerting for unusual request volumes to file processing endpoints
• Establish baseline metrics for normal file upload patterns and alert on deviations
• Monitor system resource utilization for signs of DoS attacks through resource exhaustion
• Review authentication logs to identify sessions where file operations occur without proper authentication

How to Mitigate CVE-2026-0558

Immediate Actions Required

• Upgrade lollms to a patched version that includes commit a6625dc83786ff21d109b0d545ca61b770607ef3
• If upgrading is not immediately possible, restrict network access to the /api/files/extract-text endpoint using firewall rules or reverse proxy configuration
• Implement rate limiting on the vulnerable endpoint to mitigate DoS risk
• Review access logs to determine if the vulnerability has been exploited

Patch Information

The vulnerability has been addressed in the lollms repository through commit a6625dc83786ff21d109b0d545ca61b770607ef3. This patch adds the required authentication dependencies to the file endpoints, ensuring that only authenticated users can access the text extraction functionality. Organizations should update to the latest version of lollms that includes this security fix.

For additional technical details about this vulnerability, refer to the Huntr vulnerability disclosure.

Workarounds

• Block unauthenticated access to /api/files/extract-text using a reverse proxy or web application firewall
• Disable the text extraction endpoint entirely if not required for business operations
• Implement IP-based access controls to limit endpoint access to trusted networks only
• Add rate limiting rules to prevent resource exhaustion attacks while awaiting a permanent fix

# Example nginx configuration to block unauthenticated access
location /api/files/extract-text {
    # Require authentication header
    if ($http_authorization = "") {
        return 401;
    }
    proxy_pass http://lollms_backend;
}