Abstract:

Abstract Intrusion detection and classification is one of the most essential and challenging process in the Wireless Sensor Network (WSN). Typically, the wireless networks are highly susceptible to different types of network attacks, because which reduces the lifetime of entire network by interrupting the data transmission and communication operations. Hence, the conventional works intends to develop an efficient Intrusion Detection System (IDS) frameworks by using the optimization and classification methodologies. Still, it facing the problems of high complexity in operations, more time for data training, high error rate, and inefficient detection. So, this research work objects to develop an intelligent and advanced IDS framework by implementing the novel optimization based classification methodologies. For this purpose, a hybrid Artificial Fish Integrated Particle Swarm Optimization (AFIPSO) mechanism is deployed to optimally select the features for training the data models of classifier. Consequently, the Random Artificial Neural Network Integrated Gradient Descent (RANN-GD) is implemented for accurately spotting the intrusions from the given IDS datasets based on the optimal number of features. For testing and validation, three different and emergent IDS datasets such as NSL-KDD, UNSW-NB 15 and WSN-DS have been utilized in this work. During evaluation, the performance of both existing and proposed techniques are validated and compared by using various performance measures.

Keywords: