Brittle Features May Help Anomaly Detection
This work addresses anomaly detection in security screening, offering incremental improvements in representation selection.
The paper tackled the challenge of one-class anomaly detection by evaluating representations transferred from auxiliary tasks, finding that representation choice is more critical than the detector and that adversarial brittleness correlates with performance. They achieved 96.4% anomaly detection on an X-ray security dataset, outperforming prior results.
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.