Data/ML/AISecurity
Anomaly Detection
Also written as Outlier Detection, Fraud Detection, Novelty Detection
Finding the rare cases that do not look like everything else — fraudulent payments, failing hardware, unusual logins — usually without many labelled examples of the bad ones.
Think of it like
Noticing the one engine that sounds different, when nobody has ever recorded what failure sounds like.
Junior or senior?
The hard part is the imbalance: real cases are so rare that accuracy is meaningless.
Senior sounds like
Talks about the false-alarm burden on whoever reviews the alerts.
Ask them
“How rare were the real cases, and who had to work through the false alarms?”