New ENCFORGE Ransomware Targets AI Model Files in Langflow RCE Attack

Key Points:

• A new ransomware, ENCFORGE, is targeting AI model files through a remote code execution (RCE) vulnerability in Langflow (CVE-2025-3248), which allows unauthorized access to execute arbitrary Python code.
• The impact is significant, as the ransomware encrypts critical AI infrastructure files, potentially costing organizations between $75,000 and $500,000 to rebuild affected models. The attack can compromise multiple model variants stored on the same filesystem.
• Immediate actions include patching Langflow to version 1.3.0 or later to close the RCE vulnerability and monitoring for signs of ENCFORGE activity.

Technical Details: CVE-2025-3248 has a CVSS score of 9.8 and allows remote attackers to execute arbitrary code due to unauthenticated access. The ENCFORGE ransomware uses AES-256-CTR for encryption and embeds an RSA-2048 public key for key management.

MITRE ATT&CK Techniques:

• T1203 - Exploit Public-Facing Application (Initial Access)
• T1059.001 - Command and Scripting Interpreter: Python (Execution)
• T1486 - Data Encrypted for Impact (Impact)

IOCs Mentioned:

• CVE-2025-3248
• Binary SHA-256 (packed): 8cb0c223b018cecef1d990ec81c67b826eb3c30d54f06193cf69969e9a8baea2
• Binary SHA-256 (unpacked): ea7822eac6cecef7746c606b862b4d3034856caf754c4cf69533662637905328
• Ransom note email: e78393397@proton.me