سعید طلعتیان آزاد

The availability of the system is one of the main requirements of a multimedia-based e-health application that carries critical patient health information in the network environment. On the other hand, the Distributed Denial of Service (DDoS) attack is one of the most common attacks on the availability of computer networks which can be devastating for a healthcare system. Therefore, a countermeasure to this attack has to be performed in the early steps of the attack to protect the systems against its damages. Detection methods cannot support this and are only able to detect the attack after it happened. Thus, it is necessary to predict DDoS attacks according to the evidence which the attack makes in the network in the early steps of the attack. Therefore, Prediction approaches can reduce the cost of the attacks compared to detection approaches. In this paper, we propose a new method for prediction of DDoS attack based on Lyapunov Exponent Analysis and Echo State Network (LEAESN). In this method, the future traffic of the network is predicted using the Exponential Smoothing technique, then the time series of the prediction error is calculated based on the difference of this prediction and the observed traffic of the network. As shown in this paper, this time series is chaotic in the presence of attack traffics. To predict the DDoS attack, this time series is predicted using a Recurrent Neural Echo State Network (SCESN), and the attack is detected using Lyapunov exponent analysis on the predicted time series. For the evaluation of LEAESN, we test the method on the Darpa98 dataset which consists of a standard dataset for evaluation of intrusion detection systems. LEAESN has an appropriate ability to predict the DDoS attack.