vLLM anfällig für Server-Side Request Forgery (SSRF) über MediaConnector

The SSRF vulnerability in vLLM allows an attacker to manipulate the server into making requests to unintended locations. By exploiting the differing interpretations of backslashes in Python parsing libraries used by the loadfromurl and loadfromurl_async methods, an attacker can bypass the intended host restriction. This enables access to internal services and data that should be inaccessible from the outside. The potential impact includes exfiltration of sensitive data, unauthorized access to internal systems, and potentially even remote code execution if internal services are vulnerable.

Organizations deploying vLLM for multimodal applications, particularly those with internal services accessible over the network, are at risk. Environments utilizing older versions of vLLM (≤0.9.2) without proper network segmentation or WAF protection are especially vulnerable.

• python / server:

import os
import requests

def check_ssrf(url):
    try:
        response = requests.get(url, timeout=5)
        print(f"URL {url} responded with status code: {response.status_code}")
    except requests.exceptions.RequestException as e:
        print(f"Error accessing {url}: {e}")

# Example usage - check for internal resource access
check_ssrf("http://localhost:8080/admin")

1. 2026-01-28Disclosure

   disclosure

1. Reserviert2026-01-26
2. Veröffentlicht2026-01-28
3. Geändert2026-03-27
4. EPSS aktualisiert2026-03-23

The primary mitigation for CVE-2026-24779 is to upgrade to vLLM version 0.14.1 or later, which contains the fix. If an immediate upgrade is not feasible, consider implementing network segmentation to restrict the vLLM server's access to internal resources. Additionally, implement a Web Application Firewall (WAF) with SSRF protection rules to filter outbound requests and block malicious URLs. Carefully review and validate all user-provided URLs before processing them within the MediaConnector class.