Sensor nodes constitute inexpensive, disposable devices that are often scattered in harsh environments of interest so as to collect and communicate desired measurements of monitored quantities. Due to the commodity hardware used in the construction of sensor nodes, the readings of sensors are frequently tainted with outliers. Given the presence of outliers, decision making in sensor networks becomes much harder. In this work, we introduce PAO, a framework that can reliably and efficiently detect outliers in wireless sensor networks. PAO significantly reduces the bandwidth consumption during the outlier detection procedure, while being able to operate over multiple window types. Moreover, our framework possesses the ability to operate in either an exact mode, or an approximate one that further reduces the communication cost, thus covering a wide variety of application requirements. Categories and Subject Descriptors H.4 [Information Systems Applications]: Miscellaneous General Terms Al...