CVE-2026-0760

9.8 CRITICAL

📋 TL;DR

CVE-2026-0760 is a critical remote code execution vulnerability in Foundation Agents MetaGPT's deserialize_message function. Attackers can exploit this without authentication by sending malicious serialized data, allowing arbitrary code execution as the service account. All systems running vulnerable versions of MetaGPT are affected.

💻 Affected Systems

Products:
  • Foundation Agents MetaGPT
Versions: All versions prior to patch
Operating Systems: All platforms running MetaGPT
Default Config Vulnerable: ⚠️ Yes
Notes: Any deployment with the vulnerable deserialize_message function exposed to network input is affected.

📦 What is this software?

⚠️ Risk & Real-World Impact

🔴

Worst Case

Complete system compromise with attacker gaining persistent access, data exfiltration, lateral movement, and deployment of ransomware or other malware.

🟠

Likely Case

Remote code execution leading to service disruption, data theft, and potential backdoor installation for future attacks.

🟢

If Mitigated

Limited impact with proper network segmentation and monitoring, potentially only service disruption if execution is contained.

🌐 Internet-Facing: HIGH
🏢 Internal Only: HIGH

🎯 Exploit Status

Public PoC: ✅ No
Weaponized: LIKELY
Unauthenticated Exploit: ⚠️ Yes
Complexity: LOW

ZDI-CAN-28121 indicates proof-of-concept exists in controlled environment. The vulnerability requires sending specially crafted serialized data to the vulnerable endpoint.

🛠️ Fix & Mitigation

✅ Official Fix

Patch Version: Check vendor advisory for specific patched version

Vendor Advisory: https://www.zerodayinitiative.com/advisories/ZDI-26-026/

Restart Required: Yes

Instructions:

1. Check current MetaGPT version
2. Apply vendor-provided patch or update to fixed version
3. Restart MetaGPT service
4. Verify patch application

🔧 Temporary Workarounds

Network Segmentation

linux

Restrict network access to MetaGPT service to only trusted sources

iptables -A INPUT -p tcp --dport [META_PORT] -s [TRUSTED_IP] -j ACCEPT
iptables -A INPUT -p tcp --dport [META_PORT] -j DROP

Input Validation Filter

all

Implement WAF or proxy to filter malicious serialized data

🧯 If You Can't Patch

  • Isolate the MetaGPT system from internet and untrusted networks
  • Implement strict network monitoring and alerting for suspicious deserialization attempts

🔍 How to Verify

Check if Vulnerable:

Check if MetaGPT version is unpatched and deserialize_message function is exposed to network input

Check Version:

Check MetaGPT configuration or package manager for version information

Verify Fix Applied:

Verify MetaGPT version matches patched version from vendor advisory and test with safe serialized data

📡 Detection & Monitoring

Log Indicators:

  • Unusual deserialization errors
  • Unexpected process execution from MetaGPT service account
  • Abnormal network connections from MetaGPT process

Network Indicators:

  • Malformed serialized data packets to MetaGPT port
  • Unexpected outbound connections from MetaGPT host

SIEM Query:

source="metagpt.logs" AND (event="deserialize_error" OR process_execution="unusual")

🔗 References

📤 Share & Export