Application of Hybridization Methods to Detect Network Attacks in Wireless Sensor Networks

The study proposes the structure of a network distributed attack detection system is two-tier: the first level provides primary analysis of individual packets and network connections by signature analysis, the second level processes aggregated network data streams using adaptive classifiers. The method of detection of anomalous network connections on the basis of hybridization of methods of computational intelligence for construction of multilevel schemes with arbitrary embedding of classifiers in each other and their passive connection in the process of analysis of the input vector is developed in the article. As a result of the signature analysis, it was found that the combination of detectors experimentally increased the detection of attacks by 1.28%. Software implementation of the proposed technique in the form of basic intelligent classifiers of network attacks can be used separately from the attack detection system in addressing the classification of attacks.