CVE-2024-0759
📋 TL;DR
This vulnerability in AnythingLLM allows authenticated users with manager or admin permissions to discover and potentially access other internal services on the same network through link scraping. It affects organizations hosting AnythingLLM internally where attackers have compromised privileged accounts. The risk is limited to internal network reconnaissance and potential lateral movement.
💻 Affected Systems
- AnythingLLM
📦 What is this software?
Anythingllm by Mintplexlabs
⚠️ Risk & Real-World Impact
Worst Case
Attacker discovers and accesses unprotected internal services, leading to data exfiltration, credential theft, or further network compromise.
Likely Case
Internal network reconnaissance revealing service IPs and potential access to poorly secured internal services.
If Mitigated
Limited to discovery of internal IPs without actual access due to proper network segmentation and service authentication.
🎯 Exploit Status
Requires privileged account compromise and internal network access. IP guessing/brute-forcing needed.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: Fixed in commit 0db6c3b2aa1787a7054ffdaba975474f122c20eb
Vendor Advisory: https://github.com/mintplex-labs/anything-llm/commit/0db6c3b2aa1787a7054ffdaba975474f122c20eb
Restart Required: Yes
Instructions:
1. Update AnythingLLM to version containing commit 0db6c3b2aa1787a7054ffdaba975474f122c20eb
2. Restart the AnythingLLM service
3. Verify the fix by testing link scraping functionality
🔧 Temporary Workarounds
Restrict privileged access
allLimit manager/admin permissions to trusted users only and implement strong authentication.
Network segmentation
allIsolate AnythingLLM from other internal services using network segmentation or firewalls.
🧯 If You Can't Patch
- Implement strict access controls for manager/admin roles and monitor privileged user activity
- Segment network to prevent AnythingLLM from accessing other internal services
🔍 How to Verify
Check if Vulnerable:
Check if your AnythingLLM version predates commit 0db6c3b2aa1787a7054ffdaba975474f122c20eb and test if authenticated users can scrape internal IPs via link collector.
Check Version:
Check git commit history or version metadata in AnythingLLM installation
Verify Fix Applied:
After update, test that authenticated users cannot use link collector to access internal IP addresses.
📡 Detection & Monitoring
Log Indicators:
- Unusual link scraping activity by privileged users
- Multiple failed attempts to access internal IPs
Network Indicators:
- Unusual outbound connections from AnythingLLM to internal IP ranges
- Port scanning patterns from AnythingLLM host
SIEM Query:
source="anythingllm" AND (event="link_scrape" OR event="url_collect") AND (dest_ip=10.* OR dest_ip=172.16.* OR dest_ip=192.168.*)
🔗 References
- https://github.com/mintplex-labs/anything-llm/commit/0db6c3b2aa1787a7054ffdaba975474f122c20eb
- https://huntr.com/bounties/9a978edd-ac94-41fc-8e3e-c35441bdd12b
- https://github.com/mintplex-labs/anything-llm/commit/0db6c3b2aa1787a7054ffdaba975474f122c20eb
- https://huntr.com/bounties/9a978edd-ac94-41fc-8e3e-c35441bdd12b