Anomaly detection and foresight response strategy for wireless sensor networks

Abstract

Anomaly detection is an important challenge in wireless sensor networks (WSNs) for fault diagnosis and intrusion detection applications. Sensor nodes are usually designed to be small and inexpensive, so they have limited capabilities, such as limited computational power, memory and energy. This paper presents novel light-weight distributed anomaly detection and a foresight response strategy based on support vector data description (SVDD) for wireless sensor network. SVDD could sometimes generate such a loose decision boundary, when some noisy samples (outliers) exist in the training set. In addition, it requires the solution of a computationally intensive quadratic programming approach which is not applicable in WSNs. Hence, we modified the standard version of SVDD, and proposed the Linear Programming-based Fuzzy-Constraint SVDD (LP-FCSVDD) method to detect the outliers with more accuracy in acceptable time. Then we present a foresight response strategy to resist the intentional, unintentional and false anomalies. The overall experiments show prominence of our proposed method to achieve high detection accuracies on a variety of real and synthetic wireless sensor network datasets.