Brittle Features May Help Anomaly Detection

TL;DR

This paper investigates how the choice of representations, especially brittle features transferred from auxiliary tasks, impacts one-class anomaly detection performance, showing that brittle features can enhance detection accuracy.

Contribution

It demonstrates that representation quality, particularly brittleness, is crucial for anomaly detection, and that knowledge distillation can improve detection over direct use of representations.

Findings

Representation choice outweighs detector type in importance.

Brittle features correlate with better anomaly detection.

Achieved 96.4% detection rate on X-ray security dataset.

Abstract

One-class anomaly detection is challenging. A representation that clearly distinguishes anomalies from normal data is ideal, but arriving at this representation is difficult since only normal data is available at training time. We examine the performance of representations, transferred from auxiliary tasks, for anomaly detection. Our results suggest that the choice of representation is more important than the anomaly detector used with these representations, although knowledge distillation can work better than using the representations directly. In addition, separability between anomalies and normal data is important but not the sole factor for a good representation, as anomaly detection performance is also correlated with more adversarially brittle features in the representation space. Finally, we show our configuration can detect 96.4% of anomalies in a genuine X-ray security dataset, outperforming previous results.