History-based anomaly detector: an adversarial approach to anomaly detection

Chatillon, Pierrick; Ballester, Coloma

doi:https://doi.org/10.1007/978-3-030-55180-3_58

History-based anomaly detector: an adversarial approach to anomaly detection

Citation

Chatillon P, Ballester C. History-based anomaly detector: an adversarial approach to anomaly detection. In: Arai K, Kapoor S, Bhatia R, editors. Intelligent Systems and Applications. Proceedings of the 2020 Intelligent Systems Conference (IntelliSys). Volume 1. 2020 Sept 3-4. Cham: Springer; 2020. p. 761-76. (AISC; no. 1250). DOI: 10.1007/978-3-030-55180-3_58

Abstract

Anomaly detection is a difficult problem in many areas and has recently been subject to a lot of attention. Classifying unseen data as anomalous is a challenging matter. Latest proposed methods rely on Generative Adversarial Networks (GANs) to estimate the normal data distribution, and produce an anomaly score prediction for any given data. In this article, we propose a simple yet new adversarial method to tackle this problem, denoted as History-based anomaly detector (HistoryAD). It consists of a self-supervised model, trained to recognize ‘normal’ samples by comparing them to samples based on the training history of a previously trained GAN. Quantitative and qualitative results are presented evaluating its performance. We also present a comparison to several state-of-the-art methods for anomaly detection showing that our proposal achieves top-tier results on several datasets.