Detection of DDoS Attacks based on Network Traffic Prediction and Chaos Theory

Abstract- A distributed denial-of-service (DDoS) attack is one in which a large number of compromised systems attack a single machine, thereby service is getting denied for users of the targeted system. The flood of incoming packets to the target machine essentially forces it to shut down, thus the service to the legitimate users is denied. The distributed denial of service (DDoS) attacks on computer networks or applications are facilitated by Botnet mechanisms. It is very difficult to differentiate DDoS attack traffic and legitimate traffic. Even though various approaches and systems have been proposed to detect, Distributed denial-of-service (DDoS) flooding attacks still pose great threats to the Internet. This paper proposes a DDoS detection method. Firstly network traffic is pre-processed by cumulatively averaging with a time range. Then a simple linear Auto-Regressive model is used to predict the network traffic. Secondly, Chaos theory and Network Anomaly Detection Algorithm (NADA) are used to analyse the network traffic and detect the abnormal traffic respectively. Here an assumption is made such that prediction error behaves chaotically. In-order to improve accuracy, trained neural network is used. This proposed DDoS detection algorithm effectively detect DDoS attack