Minimizing False Alarms on Intrusion Detection for Wireless Sensor Networks in Realistic Environments

Abstract

Wireless sensor networks are expected to enhance the efficiency and to reduce the cost of target detection in area surveillance systems. In order to provide accurate reports for target detection and tracking in realistic environments, not only false alarms but also the impact of weather, terrain, and ground conditions on sensor readings should be taken into account. This paper addresses how to determine a false alarm threshold dynamically and efficiently in order to minimize the false alarm probability and to maximize the probability that no target passes through without being detected. In the proposed dynamic threshold scheme, the threshold changes in accordance with the false alarm rate. This results in a better detection probability and reduces the number of false alarms. The paper proposes to reduce the impact of noise by taking a weighted average of different sensing units' readings for the same target. The fact that sensing units of different types are affected at varying degrees by the environmental factors is exploited here. In addition to analytically characterizing false alarm rate and the role of reputation values, we provide simulation results to show the improvement on the target detection accuracy by the proposed scheme. A real world target detection case is considered and the false alarm probability is reduced by 25% when compared to a single sensor reading and by 17% when compared to an non-weighted averaged reading.