STATISTICAL BASED OUTLIER DETECTION IN DATA AGGREGATION FOR WIRELESS SENSOR NETWORKS
Usman Barakkath Nisha, N. Umamaheswari, R. Venkatesh, R. Yasir Abdullah · 2014
Inconsistent data caused by compromised nodes in Wireless Sensor Networks can be detached to improve data reliability, accuracy and to make effective an d correct decisions. Multivariate Outliers normally describe the data behavior abnormality. Data aggreg ation is frequently used for the reduction of communication overhead and energy expenditure of sensor nodes during the process of data collection in Wireless Sensor Networks and also to improve the li fetime of the WSN. For the delivery of accuracy in base station, the outlier detection protocol must b e incorporated with secure data aggregation. Aggreg ation will also try to increase the circle of knowledge a nd the level of accuracy. In this paper we use mult ivariate data analysis technique, data to handle outlier in correlated variables. To achieve the reliability an d accuracy, a two phase algorithm is proposed. First, to build up a well conditioned PCA model for fault detection. Second, we use various statistical techn iques to determine similarity between the sensed da ta against the real data set. We have evaluated our al gorithm based on synthetic and real data injected w ith synthetic faults collected from a WSN. Our results concludes that the proposed algorithm achieves high true alarm rate and low false alarm rate and outper forms all the existing methods in terms of data acc uracy and reliability.