Date and time: 11.30am - 12.30pm, Friday 15th May, 2009
Venue: 10.08.04 (Building 10, Level 8, Room 4)
Abstract:
Identifying misbehaviors is an important challenge for monitoring, fault diagnosis and intrusion detection in wireless sensor networks. A key problem is how to minimize the communication overhead and energy consumption in the network when identifying misbehaviors. We treat this as a problem of distributed unsupervised learning, where the aim is to build and combine compact representations of normal behaviour based on the local measurements from each sensor. These models can be based on hyperellipsoidal, cluster-based or kernel-based representations. A key objective is to minimize the communication overhead required to share these models of normal behaviour between sensor nodes. We demonstrate on data from real-life sensor networks that our scheme achieves comparable accuracy compared to equivalent centralized approaches while achieving a significant reduction in communication overhead.
About the speaker:
Dr Chris Leckie is an Associate
Professor in the Department of Computer Science and Software Engineering at the
University of Melbourne, Australia. He has made
numerous theoretical contributions to the use of clustering for problems such
as anomaly detection in wireless sensor networks and the Internet. In
particular, he has developed efficient clustering techniques that are
specifically designed to cope with
high-dimensional and time-varying data streams, which are a major challenge in network intrusion detection. His work on filtering denial-of-service attacks on the Internet has been commercialized with an Australian company (IntelliGuard I.T.), leading to a commercial product. His research has been published in leading journals and conferences such as ACM Computing Surveys, IEEE TKDE, Artificial Intelligence, IJCAI and ICML.
Seminar Organisation
Seminars are free and open to the general public. No booking is necessary. If you are interested in giving a presentation in this seminar series, or to make suggestions for speakers, please contact Xiaodong Li, the seminar co-ordinator.