A system design for multi-phase, hybrid DDoS detection

A dominant form of cyber attack is the distributed denial of service (DDoS), which involves routing superfluous traffic to a server from multiple sources to exhaust its resources. Mudit Rathore and Abhishek Vaish at IIIT Allahabad propose a detection system for DDoS attacks. Packet filtering is used, based on the legitimacy of the origin of incoming requests, and suspicious requests are identified based on a machine learning algorithm. Features are extracted at the target end and the possibility of an attack is determined. The widespread use of the Internet has made cyber attacks a global issue.1 A dominant form of such attacks is distributed denial of service attacks (DDoS), which involves routing superfluous traffic to a server from multiple sources, making the target incapable of serving requests from legitimate users. A sophisticated and integrated mechanism is required to distinguish malicious data flow from the legitimate data flow, and differentiate between genuine and malicious sources because of the many-to-one dimension of the attack. © 2020 Elsevier Ltd