Self-Immunizing Zero-Shot Log Anomaly Detection via Log-Domain Adversarial Reasoning Simulation
We present Self-Immunizing Zero-Shot Log Anomaly Detection via Log-Domain Adversarial Reasoning Simulation, a full-stack approach for zero-shot log anomaly detection that is designed to remain reliable under realistic perturbations rather than only under clean evaluation. The central idea is to separate raw task evi...