Implementation of swarm optimized artificial neural network for network intruder detection and attack classification

Abstract

In the current era of technological advancement, the usage of computers and internet has increased drastically in the everyday life. With many crucial services operating online, the risk of cyber-attacks has made the use of online security measures inevitable. Identification of the type of the attack is an important task in safeguarding the system. Thus researchers started developing a solution for this problem and an Intruder Detection System (IDS) was created. IDS identify the suspicious behaviour in the network and improve the integrity, reliability and robustness of the host system. The IDS inspect the parameters of the system to detect any intrusion. This paper proposes an IDS using swarm optimization enhanced Artificial Neural Network (ANN) for improved performance. In this paper, KDDCUP dataset and a free dataset from kagge.com have been used for experimentation. First dimensionality reduction is applied using Principal Component Analysis (PCA). This data is fed to the Swarm Optimized ANN (SOANN) for classification. The ANN uses grey wolf optimization algorithm to optimize the weights over multiple iterations. The accuracy of the system is considered as the cost function. This cost function is implemented as the stopping criteria to terminate the optimization algorithm upon reaching the maximum accuracy and the obtained weights are used to train the final system. The performance of the system is compared against various other algorithms and has yielded better results.