How to do this? Here one algorithm: kNN Goal: Generate an anomaly score (high value->highly anomalous) Here: The anomaly score is the distance to the kth closest samples