A Framework Using Markov-Bayes’ Model for Intrusion Detection in Wireless Sensor Network

The existing Intrusion Detection Systems which are mostly designed for detecting the particular form of Intrusion for Wireless Sensor Network (WSN) has many restrictions for different types of attacks and network structures. A novel strategy for intrusion detection based on knowledge has to be applied where the attacks are prevented from creating deviation of normal features and also from various other aggregated shapes. Multiple types of attacks should be detected over different structures of network. Every year, security is provided to the networks such that intrusions are prevented, spending around more than billions of dollars, all over the world. Among them, few disruptions that occur for vital systems are considered as the most serious type of threat mainly in areas like hospitals, military, banks and other critical applications. Firstly, clusters are discovered in the features of the network using mean shift unsupervised clustering algorithm, in the phase of training. In the next stage, the revealed clusters are generalized as an anomaly, if there is definite amount of deviation taken at the preliminary stage from normal cluster which are captured during the initial stage of training, with occurrence of no attacks. Thus, as samples are traced in different stages, the threats need to be averted, with many solutions possible, and one of the best solutions possible is to design a model of Intrusion Detection System (IDS), with the approach of Bayesian and Hidden Markov Network. In the proposed framework, the IDS are designed with different processing levels of training and testing based on connection records.