An Efficient Algorithm for Anomaly Detection in Wireless Sensor Networks

Wireless sensor networks have attained remarkable attention for the past few years. They may be dropped in the real world in order to make the local measurements with the environmental condition like temperature or pressure. WSNs are exposed to faults and awful attacks due to its high density. Likewise, the sensor reading that drastically deviates from normal behavior are inaccurate and unreliable. Those abnormal data are considered as outlier, which is vulnerable to WSN and affects the data accuracy. Improper identification of outlier leads data inaccuracy and high energy consumption due to unwanted data transmission in the network. To detect outlier and improve accuracy, an algorithm with two phases is proposed. First, clustering technique describes the grouping of sensor data in training phase (Micro clustering, merging). Second, a robust density based outlier detection technique detects outlier with high accuracy. The experimental result shows that the proposed technique is having 99.56 % accuracy with low false alarm rate.