Abstract
Anomaly detection is an important problem that has been well-studied withindiverse research areas and application domains. The aim of this survey istwo-fold, firstly we present a structured and comprehensive overview ofresearch methods in deep learning-based anomaly detection. Furthermore, wereview the adoption of these methods for anomaly across various applicationdomains and assess their effectiveness. We have grouped state-of-the-artresearch techniques into different categories based on the underlyingassumptions and approach adopted. Within each category we outline the basicanomaly detection technique, along with its variants and present keyassumptions, to differentiate between normal and anomalous behavior. For eachcategory, we present we also present the advantages and limitations and discussthe computational complexity of the techniques in real application domains.Finally, we outline open issues in research and challenges faced while adoptingthese techniques.