Description

Abstract

We proposed a distributed detection system for clustered wireless sensor networks. The proposed

distributed learning algorithm for the training of SVM in WSN reaches high accuracy for

detecting the normal and anomalous behaviour (accuracy rate over 98%). Also a combination

between the SVM classifier and Signature Based Detection achieve a high detection rate with low

false positive rate. Communication in WSN consumes a high energy, as an example one bit

transmitted in WSNs consumes about as much power as executing 800-1000 instructions. The

training process is carried out with IDS nodes. These nodes need to compute and transmit only a

set of data vector (support vector) between each others, instead of transmitting all captured data to

a centralized point, then train a SVM classifier.