Analysis of Anomalies in IBRL Data from a Wireless Sensor Network Deployment

Abstract

Detecting interesting events and anomalous behaviors in wireless sensor networks is an important challenge for tasks such as monitoring applications, fault diagnosis and intrusion detection. A key problem is to define and detect those anomalies with few false alarms while preserving the limited energy in the sensor network. In this paper, using concepts from statistics, we perform an analysis of a subset of the data gathered from a real sensor network deployment at the Intel Berkeley Research Laboratory (IBRL) in the USA, and provide a formal definition for anomalies in the IBRL data. By providing a formal definition for anomalies in this publicly available data set, we aim to provide a benchmark for evaluating anomaly detection techniques. We also discuss some open problems in detecting anomalies in energy constrained wireless sensor networks.