Outlier Detection Using Support Vector Machine in Wireless Sensor Network Real Time Data

Abstract

Outlier detection has many important applications in sensor networks, e.g., abnormal event detection, animal behavior change, intruder detection etc. Outliers in wireless sensor networks (WSNs) are sensor nodes that issue attacks by abnormal behaviours and fake message broadcasting. The probable sources of outliers include noise and errors, events, and malicious attacks on the network. Wireless sensor networks (WSNs) are more likely to generate outliers due to their special characteristics, e.g. constrained available resources, frequent physical failure, and often harsh deployment area. In this project we motivate our technique in the context of the problem of outlier detection. This paper is going to present the real time network outlier detection method in the wireless sensor networks. We proposed the technique to classify the sensor node data as local outlier or cluster outlier or network outlier using Standard Support Vector Machine classification method which is one of the best classification methods among the various outlier detection methods. If the data is classified as network outlier then it may be due an event otherwise if it is classified as a cluster outlier then it is an error in the cluster due to some environmental factor or network otherwise is an error in the sensor node due to some defect in that sensor. Experiments with real data show that our method is efficient and accurate to detect the outliers in real time. The real time data are collected from the Sensor Scope system and implemented using MATLAB.