Perfect contextual information privacy in WSNs under colluding eavesdroppers

We address the problem of preserving contextual information privacy in wireless sensor networks (WSNs). We consider an adversarial network of colluding eavesdroppers that are placed at unknown locations. Eavesdroppers use communication attributes of interest such as packet sizes, interpacket timings, and unencrypted headers to infer contextual information, including the time and location of events reported by sensors, the sink's position, and the event type. We propose a traffic normalization technique that employs a minimum backbone set of sensors to decorrelate the observable traffic patterns from the real ones. Compared to previous works, our method significantly reduces the communication overhead for normalizing traffic patterns.