CVE-2024-3271
📋 TL;DR
A command injection vulnerability in the run-llama/llama_index repository allows attackers to bypass security checks and execute arbitrary code on servers. This remote code execution vulnerability affects any system using the vulnerable safe_eval function. Users of llama_index who process untrusted input are at risk.
💻 Affected Systems
- run-llama/llama_index
📦 What is this software?
Llamaindex by Llamaindex
⚠️ Risk & Real-World Impact
Worst Case
Complete server compromise allowing data theft, lateral movement, and persistent backdoor installation
Likely Case
Unauthorized code execution leading to data exfiltration, service disruption, or cryptocurrency mining
If Mitigated
Limited impact with proper network segmentation and least privilege controls
🎯 Exploit Status
Exploit bypasses underscore check by using alternative command injection techniques
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: Commit 5fbcb5a8b9f20f81b791c7fc8849e352613ab475 and later
Vendor Advisory: https://github.com/run-llama/llama_index/commit/5fbcb5a8b9f20f81b791c7fc8849e352613ab475
Restart Required: No
Instructions:
1. Update llama_index to latest version. 2. Verify commit hash includes 5fbcb5a8b9f20f81b791c7fc8849e352613ab475. 3. Test safe_eval function with malicious inputs.
🔧 Temporary Workarounds
Disable safe_eval function
allTemporarily disable or replace the vulnerable safe_eval function with a secure alternative
# Modify code to bypass safe_eval or implement input validation
Input sanitization wrapper
allAdd additional input validation before passing to safe_eval
# Implement regex filtering for dangerous characters and patterns
🧯 If You Can't Patch
- Network segmentation to isolate vulnerable systems from critical assets
- Implement strict input validation and output encoding for all user-supplied data
🔍 How to Verify
Check if Vulnerable:
Check if your code uses safe_eval function from llama_index and verify versionCheck Version:
git log --oneline | grep 5fbcb5a8b9f20f81b791c7fc8849e352613ab475Verify Fix Applied:
Test safe_eval with crafted inputs containing command injection attempts without underscores📡 Detection & Monitoring
Log Indicators:
- Unusual process spawns from Python applications
- Unexpected system command execution patterns
Network Indicators:
- Outbound connections from application servers to unusual destinations
- Command and control traffic patterns
SIEM Query:
source="application.logs" AND ("safe_eval" OR "llama_index") AND (process_execution OR command_injection)🔗 References
- https://github.com/run-llama/llama_index/commit/5fbcb5a8b9f20f81b791c7fc8849e352613ab475
- https://huntr.com/bounties/9b32490e-7cf9-470e-8d49-ba083ae7a279
- https://github.com/run-llama/llama_index/commit/5fbcb5a8b9f20f81b791c7fc8849e352613ab475
- https://huntr.com/bounties/9b32490e-7cf9-470e-8d49-ba083ae7a279
