An Anomaly Detection Algorithm Based on Real-Time Node Flow
Qingping Wang · Piezoelectrics and Acoustooptics · 2011
Sensor nodes have specific properties such as lack of mobility and relatively predictable traffic patterns which can not allow for detection of anomalies in their networking behavior.In this paper we introduce a real-time,node-based anomaly detection algorithm that observes the arrival processes experienced by a sensor node.A new arrival model for the traffic is developed which can be received by a sensor node and devise a scheme to detect anomalous changes in this arrival process.The detection algorithm keeps short-term dynamic statistics using a multi-level,sliding window event storage scheme.In order to detect the anomalies for the wireless sensor networks,arrival processes at different time scales are compared using node resource computable,low-complexity,aggregate features.