Lightweight Location Verification Algorithms for Wireless Sensor Networks

Abstract

The knowledge of sensors' locations is crucial information for many applications in Wireless Sensor Networks (WSNs). When sensor nodes are deployed in hostile environments, the localization schemes are vulnerable to various attacks, e.g., wormhole attack, pollution attack, range enlargement/reduction attack, and etc. Therefore, sensors' locations are not trustworthy and need to be verified before they can be used by location-based applications. Previous verification schemes either require group-based deployment knowledge of the sensor field, or depend on expensive or dedicated hardware, thus they cannot be used for low-cost sensor networks. In this paper, we propose a lightweight location verification system that performs both “on-spot” and “in-region” location verifications. The on-spot verification intends to verify whether the locations claimed by sensors are far from their true spots beyond a certain distance. We propose two algorithms that detect abnormal locations by exploring the inconsistencies between sensors' claimed locations and their neighborhood observations. The in-region verification verifies whether a sensor is inside an application-specific verification region. Compared to on-spot verification, the in-region verification is tolerable to large errors as long as the locations of sensors don't cause the application to malfunction. We study how to derive the verification region for different applications and design a probabilistic algorithm to compute in-region confidence for each sensor. Experiment results show that our on-spot and in-region algorithms can verify sensors' locations with high detection rate and low false positive rate. They are robust in the presence of malicious attacks that are launched during the verification process. Moreover, compared with previous verification schemes, our algorithms are effective and lightweight because they do not rely on the knowledge of deployment of sensors, and they don't require expensive or dedicated hardware, so our algorithms can be used in any low-cost sensor networks.