Abstract — The evolution of Wireless Sensor networks (WSNs)

is a demandable, efficient and emerging area of Computer Science Engineering which has been currently employed in various field of engineering particularly in communication system to make it effective and reliable. For successful application of WSN it is important to maintain the basic security level, both from external and internal attacks else entire network may collapse. In the field of WSNs, if the sensed patterns that significantly deviate from the normal pattern are considered as outlier. The possible source of outlier includes noise, errors, events and malicious attack on the network. WSNs are more likely to generate outlier due to their special characteristics like constrained available with the resources causing frequent physical failure and harsh deployment area. In this paper we have investigated and experimented different outlier attack in WSN and how these attacks are efficiently detected and rectified. The usual outlier detection techniques are not directly applicable to wireless sensor network due to the nature of sensed data, specific requirements and limitations of the WSNs. Our proposed method will train and test data set for detection as well as rectification of outlier due the malicious attack and identify the affected sensor nodes in a largely deployed cluster based wireless sensor network under a common outlier detection framework which is designed by some well specified supervised learning and classification based data mining technique