This is published under the terms of CC BY-NC.

Downloaded: 788 times

Abstract

This paper presents a system implementation to detect and classify different DDoS attacks. The system adopts features of inter-arrival time, entropy, and packet length distribution for a hybrid machine learning model, which is based on the hierarchical temporal memory (HTM) with a k-nearest neighbors (KNN) classifier that can mine network traffic anomalies. Furthermore, it can incrementally learn new traffic behavior and recognize new types of attacks. Finally, system evaluation is conducted based on the CICDDoS 2019 dataset. Thus, the proposed system can successfully identify different attacks with high detection rate, accuracy, and precision.