Anomaly Detection for Big Data Using Efficient Techniques: A Review

Abstract

There has been a trending impacting application in different sectors such as hospitals, banking, defense, retail and social networks, the rate at which information needs to be kept safe has been a major concern. The security being compromised at a rapid rate which is shown through the increasing rise in frauds has caused a serious impact. Anomaly detection has affected huge sectors and made its way in various applications including detect of fraudulence records, patterns, behaviors, third-party attack or intruder detection in networks and cyber-security. A great deal of research is being carried out in these areas, which uses various approaches such as statistical, data analytics, machine learning and so forth from which anomalies could be extracted. The interdependencies between data can be depicted and represented using graphs in an effective way. However, few works have focused on detecting anomalous behaviors, patterns, links and points in data represented as graphs with multi-dimension especially for detecting outliers in unstructured data. Indeed, data represented as graphs show interdependencies and should be taken into account while detecting anomalies. In this paper, a survey is carried out on the static, dynamic and machine learning approaches for detecting anomalies in the graph data structures. In addition, more focus is toward efficient graph-based techniques that can help in anomaly detection in big data transactions which are presented. Along the way, few open issues and research directions are also mentioned.