analyzing-linux-audit-logs-for-intrusion

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow reads outsider-authored free text from the local audit log file (e.g., /var/log/audit/audit.log) and parses its raw lines into structured fields that are then included in the agent’s LLM context via printed JSON/summary output; audit log contents can include attacker-controlled strings (e.g., command/exe/path) and are not authored by the operating user.