GAN-based anomaly detection: A review

Supervised learning algorithms have shown limited use in the field of anomaly detection due to the unpredictability and difficulty in acquiring abnormal samples. In recent years, unsupervised or semi -supervised anomaly-detection algorithms have become more widely used in anomaly-detection tasks. As a form of unsupervised learning algorithm, generative adversarial networks (GAN/GANs) have been widely used in anomaly detection because GAN can make abnormal inferences using adversarial learning of the representation of samples. To provide inspiration for the research of GAN-based anomaly detection, this review reconsiders the concept of anomaly, provides three criteria for discussing the anomaly detec-tion task, and discusses the current challenges of anomaly detection. For the existing works, this review focuses on the theoretical and technological evolution, theoretical basis, applicable tasks, and practical application of GAN-based anomaly detection. This review also addresses the current internal and external outstanding issues encountered by GAN-based anomaly detection and predicts and analyzes several future research directions in detail. This review summarizes more than 330 references related to GAN-based anomaly detection and provides detailed technical information for researchers who are interested in GANs and want to apply them to anomaly-detection tasks. (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

This review explores the application of generative adversarial networks (GANs) in anomaly detection, discussing the concept of anomalies, criteria for anomaly detection tasks, and analyzing current challenges and future research directions.