Isolation Forest
A tree-based anomaly detector scoring points by how few random splits isolate them, with short paths meaning anomalous and long paths meaning typical.
Each isolation tree splits data with random cuts at random feature values, and a point in a sparse region gets isolated in its own leaf after just a few cuts, while a point in a dense region needs many more. Averaging path length across many independently randomized trees corrects for any single tree's lucky or unlucky split.
Unlike distance- or density-based detectors, no distance metric is ever computed, which lets it scale past the dimensionality limits those methods run into.