Identification of Intruder Paths in Sensor Networks

• Activity Number: 464
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): James Shine*+ and James Gentle
• Companies: U.S. Army Topographic Engineering Center and George Mason University
• Keywords: network analysis ; sensors ; intrusion detection ; probabilistic network

Abstract

Sensor networks are designed to discover as much information about anomalous activities (such as intrusion) as possible. Some important knowledge for such networks are normal sensor activity levels, sensor alarms (when sensors are activated more than normal), and patterns over multiple sensors. Our work focuses on methods for discovering possible paths that intruders might be taking based on a record of sensor activations. We use the distance and time between successive sensor activations. At any activation, the primary question is whether the activity is caused by a new intruder, or by one who has already activated an earlier sensor (or sensors). Our approach looks in both temporal directions to examine and weight all possible intruder paths. We are developing a tool that outputs possible paths given input from sensors of time and location. Initial results from the tool are presented.