Detection Mechanism Against DDoS Attacks based on Convolutional Neural Network in SINET

Abstract

From DoS (Distributed Denial of Service) to DDoS (Distributed Denial of Service), DoS attacks have always been an important threat to network security. In Smart Identifier network, resource adaption resolution server is vulnerable to single point failures due to DDoS attacks, which may leads to the paralysis of the whole network, thus it is particularly important to detect DDoS attacks against it. This paper proposes a detection mechanism against DDoS attacks for resource adaptation resolution server. Firstly, we locate suspicious connections according to the characteristics of DDoS attacks and then migrate the connections to the backup server. The backup server is equipped with a convolutional neural network, which can accurately determine whether the traffic on the suspicious connections is attack traffic or not. The simulation results show that the detection mechanism can effectively detect whether a DDoS attack occurs on a certain connection, migrate suspicious connections, and alleviate the pressure of network and ensure its normal operation.