Anomaly detection identifies data points that deviate from normal patterns, fraudulent transactions, failing machines, network intrusions, data-quality breaks,
Identifying data points that deviate from normal patterns, fraudulent transactions, failing machines, network intrusions, data-quality breaks, usually by learning what 'normal' looks like and flagging departures from it.
Because it needs no labeled failures: you model the baseline and alert on deviation. The craft is precision tuning, since alert fatigue from too many false positives kills more deployments than missed anomalies do.