An Information-Theoretic Approach to Active Adversaries in Networks
|Event Date:||January 23, 2017|
|Speaker:||Professor Oliver Kosut|
|Speaker Affiliation:||Arizona State University|
|Sponsor:||Prof. Xiaojun Lin|
|Time:||1:30 p.m.-2:30 p.m.
|Open To:||ACCEPTABLE FOR ECE 694
|School or Program:||Electrical and Computer Engineering
The vast and distributed nature of communication networks makes them inherently vulnerable to attack. Of particular importance are active attacks, in which a malicious intruder acts to subvert the natural behavior of the system. These attacks include distributed denial of service attacks, wireless jamming, man-in-the-middle attacks, and Byzantine attacks. This talk presents an information-theoretic framework for characterizing the fundamental limits of active attacks in networks.
Two recent lines of work within this framework will be presented. The first focuses on wireline networks subject to attacks from omniscient adversaries that take control of a subset of nodes in the network. To counteract these attacks, the new paradigm of Polytope Codes is presented, in which parity checks over the integers (rather than a finite field) are used to limit the attacker’s impact. These codes are shown to outperform standard linear codes.
The second line of work considers active attackers in the more challenging regime of noisy networks. Results are presented wherein a network containing a noisy channel in the presence to a malicious attacker is equivalent to a channel with no attacker at all. This equivalence is in the very strong sense that any strategy used with one can be used with the other, even if the optimal strategies for a given network are unknown. This constitutes a “reduction” from an adversarial problem to a non-adversarial problem. Under certain connectivity conditions on the network, this reduction is shown to be complete.
Oliver Kosut received B.S. degrees in electrical engineering and mathematics from the Massachusetts Institute of Technology, Cambridge, MA in 2004 and the Ph.D. degree in electrical and computer engineering from Cornell, Ithaca, NY in 2010. Since 2012, he has been an Assistant Professor at Arizona State University, Tempe, AZ. Previously, he was a Postdoctoral Associate in the Laboratory for Information and Decision Systems at MIT from 2010 to 2012. His research interests include information theory, cyber-security, and power systems. Prof. Kosut received the NSF CAREER award in 2015.