Wireless Motes Outlier Detection Taxonomy Using ML-Based Techniques

WSNs have enthused resurgence in research on machine learning-based approaches with the intent of overcoming the physical restraints of sensors. Although resource constrained in nature, WSNs domain has tremendous potential for building powerful applications, each with its own individual characteristics and requirements. This fascinating field of WSNs although comprises of various research issues and challenges, viz. energy efficiency, localization, etc., which needs to rectified. One such prominent challenge is detection of outliers whose chore is to preclude any kind of malicious attacks in the network or lessen the noisy error prone data in millions of wireless sensor networks. To do so, the methodologies developed needs to take care of inherent limits of sensor networks so that the energy intake of motes is minimum and lifespan of the motes is maximized. Consequently, the quality of data must be thoroughly patterned as any kind of outlier in the sensed network may degrade the quality of the data and hence affect the final decision. Thus, it becomes imperious to retain the quality of the data. Numerous ML-based methodologies have been used by several researchers over the time to detect any form of outliers or anomaly present in the network. In this paper, some machine learning methodologies have been discussed which have proved their mettle in outlier detection for sensor networks. This paper presents a brief survey on outlier detection in WSNs data using various ML-based techniques.