Comparative Analysis of Anomaly Detection Techniques Using Generative Adversarial Network

Abstract

Anomaly detection in a piece of data is a challenging task. Researchers use different approaches to classify data as anomalous. These include traditional, supervised, unsupervised, and semi-supervised techniques. A more recently introduced technique is Generative Adversarial Network (GAN), which is a deep learning-based technique. However, it is difficult to choose one anomaly detection algorithm over another because each algorithm stands out with its own performance. Therefore, this paper aims to provide a structured and comprehensive understanding of machine-learning based anomaly detection techniques. This paper carries out a survey of the existing literature on machine learning-based algorithms for anomaly detection. This paper places a special emphasis on Generative Adversarial Network-based algorithms for anomaly detection, since it is the most widely used machine-learning based algorithm for anomaly detection.