Date:  September 26, 2014

Time:  12:20-1:10pm EST

Location:  Room 404 Min H. Kao Building, Knoxville, TN

Online Streaming Link: Live Video

Title:  Enhancing Network Security Using 'Learning-From-Signals' and Fractional Fourier Transform Based RF-DNA Fingerprints

Presenter:  Dr. Mark Buckner, Group leader of the Power and Energy Systems Department of Oak Ridge National Laboratory

Abstract:  Wireless Access Points (WAP) remain one of the top 10 network security threats.  ZigBee wireless devices are essential components in smart grid, health care, industrial control, and building automation networks. However, recent attack tools such as KillerBee and Api-do are routinely used to attack these networks by impersonating legitimate devices through address spoofing and replayed commands.  Our research has focused on developing a physical (PHY) layer aware Radio Frequency (RF) air monitoring system with multi-factor authentication to provide a first-line of defense for network security--stopping attackers before they can gain access to critical infrastructure networks through vulnerable WAPs or ZigBee devices.  This talk will present results on the identification of OFDM-based 802.11a WiFi devices using RF Distinct Native Attribute (RF-DNA) fingerprints produced by the Fractional Fourier Transform (FRFT).   These fingerprints are input to a "Learning from Signals" (LFS) classifier which uses hybrid Differential Evolution/Conjugate Gradient (DECG) optimization to determine the optimal features for a low-rank model to be used for authentication.  Results are presented for devices under the most challenging conditions of intra-manufacturer classification, i.e., same-manufacturer, same-model, differing only in serial number.  The results of Fractional Fourier Domain (FRFD) RF-DNA fingerprints demonstrate significant improvement over results based on Time Domain (TD), Spectral Domain (SD) and even Wavelet Domain (WD) fingerprints.  A path for developing a real-time air monitor for wireless situational awareness and PHY-layer aware multi-factor authentication will also be presented.

Bio:  Mark Buckner is a group leader of the Power and Energy Systems department of Oak Ridge National Laboratory.  He received his Ph.D. and M.S. from University of Tennessee, Knoxville, in Nuclear Engineering, with an emphasis on Applied Artificial Intelligence.  At ORNL he develops (1) bio-inspired approaches to transduction, signal processing, learning and cognition; (2) novel machine learning approaches for “signals mining” and fingerprint/signature generation for multi-factor authentication for wireless communications and embedded electronic devices; and (3) software-defined instrumentation and cognitive control for secure, resilient, energy delivery systems.  His Primary interest is in strategic trans-disciplinary research and the development of a unified theory of computational intelligence and learning machines, e.g., “Big Learning on Big Iron with Big Data.”