High Dimension Clustering for Computer Intrusion Detection

Period of Performance: 04/13/2000 - 10/13/2000

$65K

Phase 1 SBIR

Recipient Firm

Quadrant Engineering, Inc.
107 Sunderland Road
Amherst, MA 01002
Principal Investigator

Abstract

This Phase I SBIR proposal describes high dimensional data classificationalgorithm applicable to the problem of real-time intrusion detection. Ourapproach to this problem involves using generic, robust data classificationalgorithms for very large sets of high dimensional data vectors. Thealgorithm is based on three successful projects in data clustering carriedout in recent years by researchers at the University of Massachusett. Ourphase I goal is to test our clustering algorithms on ground truth data ina mutually blind fashion and to clarify the concept of similarity used inthe particular case of intrusion detection. The algorithms are developedindependently of the ground truth data and will be generically applicable. Applications of this technology include protecting government, military and private computer systems against unauthorized intrusion.