On Data-Centric Intrusion Detection in Wireless Sensor Networks

Wireless sensor networks (WSN) are increasingly used to support critical applications - especially in enterprise settings. If the sensor data collected through the network is incorrect, such applications cannot run reliably. Thus, detecting the occurrence of abnormal sensor values is crucial. In this paper we develop three decentralized, lightweight data anomaly detection mechanisms that can be run directly on sensor nodes. These algorithms are evaluated with a real dataset to which we added plausible attacks. Further, they are compared to standard centralized anomaly detection mechanisms.