SSRF in POST /worker_generate_stream API endpoint in lm-sys/fastchat

The SSRF vulnerability in fastchat's Controller API Server poses a significant risk. An attacker can exploit this by combining the /workergeneratestream endpoint with the /register_worker endpoint to craft malicious requests. This allows them to bypass security controls and potentially access internal resources, sensitive data, or even trigger actions on other systems the controller has access to. The impact could range from data exfiltration to complete system compromise, depending on the permissions associated with the controller’s credentials. This vulnerability shares similarities with other SSRF exploits where attackers leverage internal network access to gain unauthorized access.

Organizations deploying lm-sys/fastchat for large language model serving, particularly those with exposed Controller API Servers, are at significant risk. This includes research labs, AI development teams, and any environment where fastchat is used to manage and orchestrate language models. Shared hosting environments where multiple users share the same fastchat instance are also particularly vulnerable.

• python / server: Monitor outbound network traffic from the fastchat Controller API Server for unexpected destinations. Use tools like tcpdump or Wireshark to capture and analyze traffic.

tcpdump -i any -n port 80 or port 443 | grep -i 'example.com'

• python / server: Examine fastchat logs for unusual requests to the /workergeneratestream and /register_worker endpoints. Look for requests with malformed URLs or unexpected parameters.

# Example log analysis (replace with your actual log parsing) 
import re
with open('fastchat.log', 'r') as f:
    for line in f:
        if re.search(r'/worker_generate_stream.*(http://|https://)', line):
            print(line)

• generic web: Check for unusual outbound connections from the fastchat server using ss or netstat.

ss -t -a | grep fastchat

1. 2024-12-30Disclosure

   disclosure

2. 2024-12-30Patch

   patch

1. Reserviert2024-10-16
2. Veröffentlicht2024-12-30
3. EPSS aktualisiert2026-04-02

The primary mitigation for CVE-2024-10044 is to upgrade to a patched version of fastchat as soon as it becomes available. Until a patch is applied, implement strict network access controls to limit the controller's outbound connections. This can be achieved through firewalls or network segmentation. Additionally, thoroughly validate all user-supplied input to prevent attackers from injecting malicious URLs. Consider implementing a Web Application Firewall (WAF) with SSRF protection rules to block suspicious requests. Monitor logs for unusual outbound connections originating from the controller server.