Detection Library
mediumexperimentalLinuxAI/MLNetworkT1071.001
LLM Service Connecting To Unexpected External Destination
Detects LLM service processes making outbound network connections to external destinations outside the OCI and private network baseline. A prompt injection may instruct the model to exfiltrate data or beacon to an attacker-controlled server.
Updated Jan 15, 2025 · Detection Engineering Team
llmprompt-injectionlinuxnetworkowasp-llm01
Problem Statement
Prompt injection can cause an LLM to initiate outbound connections for data exfiltration or C2 callback. Unexpected external network connections from LLM service processes are a reliable post-injection signal.
Sample Logs
{"timestamp":"2025-01-15T16:55:03Z","computer_name":"llm-host-03","user":"llm_svc","image":"/srv/llm/app/server.py","destination_hostname":"attacker.example.com","destination_ip":"198.51.100.42","destination_port":443}Required Fields
image
destination_hostname
destination_ip
destination_port
user
computer_name
False Positives
- ·LLM services with legitimate integrations to external APIs (HuggingFace, OpenAI)
- ·Telemetry or licensing beacons to approved vendor endpoints
Tuning Guidance
Maintain an allowlist of approved external destinations for each LLM service. Alert on first-seen destinations using a lookup table approach.