COLLOQUIUM
Department of Computer Science and Engineering
and
Advanced Solutions Group
University of South Carolina
 
Singular Value Decomposition as a Tool for Anomaly Detection

Victor A. Skormin

Department of Electrical Engineering
Binghamton University

Date: September 29, 2004 (Wednesday)
Time: 1-2:30
Place: Swearingen 1A03 (Faculty Lounge)

Abstract

The immune system demonstrates the most efficient mechanism capable of
recognition of “self” from “non-self” when both objects
are represented by very complex three-dimensional protein
structures. An attempt to model this process has resulted in the
approach applicable to a wide class of recognition-based problems. It
presents the means for mathematical description of complex phenomena,
and then for the generic feature extraction and the utilization of the
extracted features for a reliable and relatively simplistic
recognition procedure. The underlying mathematical engine is the
Singular Value Decomposition (SVD). It is important that the approach
facilitates both supervised and unsupervised learning facilitating
recognition.  It is known that the pattern of data traffic through for
a fixed computer network is indicative of the network status,
i.e. normal or abnormal operation caused by information attacks. The
abnormal operations could be further classified according to
particular types of attacks. Monitoring data traffic could be easily
achieved without placing unnecessary burden on the network resources;
data traffic data could be represented by vectors folded to a matrix
format and as such subjected to various analyses. It is important that
modern numerical tools can easily manipulate a 1000?1000 matrix that
is more than sufficient for many network security-related tasks.
       
Victor A. Skormin is a Professor of Electrical Engineering at the
Watson School of Engineering, Binghamton University (SUNY).  He holds
a MS (1968) from Kazakh National Technical University, Kazakhstan, and
Ph.D. (1975) degrees from the Moscow Institute of Steel and Alloys,
Russia. His area of research includes modern control theory and
applications (motion control, pointing-acquisition-tracking systems in
laser communication, novel robotics-based gimbals systems,
high-performance hybrid laser positioning systems), technical
diagnostics (system diagnostics for power generators and avionics),
mathematical modeling and system optimization, information security
(biological approach to system information security, detection of the
“gene of self-replication” in malicious codes, immunocomputing),
and biometrics (effects of intoxicants and fatigue on speech). His
current research is funded by NSF, NASA, and Air Force. Dr. Skormin is
a recipient of the IEEE Region I Award “For Leadership in
Establishing University-Industry Links in Research and Education”,
the University Award for Graduate Teaching from Binghamton University,
the SUNY Chancellor’s Award for Excellence in Teaching, the SUNY
Chancellor’s Award for Excellence in Scholarship and Creative
Activities, and the 2004 Book of the Year Award for his book on
IMMUNOCOMPUTING from the International Institute for Advanced Studies
in Systems Research and Cybernetics (IIAS).  He is a Senior Member of
the IEEE and served as the Editor for Space Systems of the IEEE
Transactions on Aerospace and Electronic Systems.  From 1999-2000
Dr. Skormin was the appointed National Research Council’s Senior
Researcher with the Air Force.  In 1999 Dr. Skormin was awarded the
title of Honorary Professor of the Kazakh National Technical
University, Almaty, Kazakhstan. In 2000 he was elected an
International Member of the Russian Academy of Navigation and
Control. He is a Member of the Cyber Security Task Force of
SUNY-Central. He authored three books and many journal papers.