tailieunhanh - Adaptive scheme for outliers detection in wireless sensor networks

The Nearest Neighbour Search is one of these techniques. It uses the Euclidean or Mahalanobis distance for detecting outliers in a given sensor networks. In this paper, we proposed a distributed adaptive approach for the detection of outliers. The proposed approach is based on Euclidean or Mahalanobis distance, depending on the size of the data collected. Extensive experiments have been conducted and the results confirmed the effectiveness of the proposal. | International Journal of Computer Networks and Communications Security VOL. 5, NO. 5, MAY 2017, 105–114 Available online at: E-ISSN 2308-9830 (Online) / ISSN 2410-0595 (Print) Adaptive Scheme for Outliers Detection in Wireless Sensor Networks Blaise Omer Yenké1, Moussa Aboubakar1, Chafiq Titouna2, 3, Ado Adamou Abba Ari2, 4 and Abdelhak Mourad Gueroui2 1 2 LASE Laboratory, University of Ngaoundéré, CAMEROON LI-PaRAD Laboratory, Université Paris Saclay, University of Versailles Saintt-Quentin-en-Yvelines, FRANCE 3 Department of Computer Science, Faculty of Mathematics and Computer Science, University of Batna 2, ALGERIA 4 Department of Mathematics and Computer Science, Faculty of Science, University of Maroua, CAMEROON ABSTRACT Nowadays, Wireless Sensor Networks (WSNs) have become a way of information gathering and monitoring. These particular kinds of ad hoc networks find applications in various fields such as healthcare, army and environment, etc. However, WSNs are subjected to a number of faults, either due to the worst communication links, the hostile environment in which they are deployed or due to energy constraints. Thereby, to ensure the Quality of Service (QoS) in such networks, it is important that WSNs become capable of detecting and recovering erroneous data. In this way, many outlier detection techniques have been proposed. These techniques are based on information theory, statistics and other techniques. The Nearest Neighbour Search is one of these techniques. It uses the Euclidean or Mahalanobis distance for detecting outliers in a given sensor networks. In this paper, we proposed a distributed adaptive approach for the detection of outliers. The proposed approach is based on Euclidean or Mahalanobis distance, depending on the size of the data collected. Extensive experiments have been conducted and the results confirmed the effectiveness of the proposal. Keywords: Wireless Sensor Networks, Outlier Detection, Mahalanobis .