READ: Reliable Event and Anomaly Detection System in Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) are widely used in a broad range of IoT applications, mostly in surveillance systems. Events like fire occurrence or terrorist intrusion are detected and propagated to concerned authorities to ensure people safety and security. Event detection systems have to be reliable to minimize false alarms and not to miss important events detection. This paper proposes Reliable Event and Anomaly Detection (READ for short) system. The proposed approach relies on three complementary building blocks: Support Vector Machines based anomaly detection components, event detection components and a reputation system. READ novel idea consists in activating its components only on-demand for energy reduction purpose. The activation task depends on nodes trust level. Based on the performed simulations results and with regard to the existing solutions, this approach maximizes the detection rate, minimizes the false positive rate while reducing energy overhead.