You are here

ECE PhD Defense: Koorosh Firouzbakht


442 Dana

August 3, 2015 11:30 am to 1:30 pm
August 3, 2015 11:30 am to 1:30 pm



Abstract: Game theory has been widely used to study many specific jamming and interactive communication problems in wireless networks. However, these theoretical studies are only applicable to very specific problems and the study is often based on a particular features or weaknesses of the communication system under study. These features and weaknesses are in most cases unrelated to the physical layer of the protocol stack. Consequently, these studies fail to accurately model more general problems or even the same problem with minor changes in the original assumptions. This limitation demands development of a more general theoretical frameworks that can be applicable to a wide range of jamming scenarios.

In this dissertation, we develop two general game-theoretic frameworks, constrained zero-sum and constrained bimatrix, that can be used to model many interactive communication scenarios in wireless networks when physical layer jamming is present.

In constrained games, players' strategies are limited to a subset of all possible strategies and as a result, a broader class of problems can be modeled by using these frameworks.

Furthermore, we formulate the interactions between adaptive communicating nodes and smart power limited adversaries by constrained zero-sum and constrained bimatrix games and provide necessary and sufficient conditions under which existence of the Nash equilibrium solutions for these non-typical games are guaranteed.

We show that constrained zero-sum games and linear programming have deep connections. For every constrained zero-sum game there exists a linear program whose solution yields a Nash equilibrium and every equilibrium solution of the zero-sum game is a solution of the linear program. Similarly, we show that there exists a similar relationship between the Nash equilibrium solutions of constrained bimatrix games and global solutions of a quadratic program. Every NE solution of the bimatrix game is a global maximizer of a quadratic program and every global maximum of the quadratic program yields a NE solution.

We use our constrained frameworks to study some typical jamming problems in packetized networks where we derive analytical as well as numerical results for the optimal strategies and closed form expressions for the expected value of the game at the the Nash equilibrium.

Our analytical results suggest that a strategic jammer that uses optimal jamming strategies can significantly degrade the performance of packetized networks compared to non-strategic jammers.. Furthermore, we prove that there exists a certain threshold on jammer's average power, $J_{\text{TH}}$, such that if the jammer's average power exceeds $J_{\text{TH} }$, the game value at the Nash equilibrium is the same as the case when the jammer uses his maximum power all the time.  

Additionally, we study the performance of an adaptive OFDM wireless communication system under power limited jamming and we show that with modest assumptions, this problem can be formulated into the constrained zero-sum or constrained bimatrix frameworks also. 

Finally, inspired by the superposition coding technique used in broadcast channels, we propose an adaptive multilayer superposition coding technique to improve the performance of packetized networks. Our analytical results shows that superposition coding not only achieves better average performance under jamming, but also increases the jamming threshold. 


Advisor: Professor Masoud Salehi

Committee Members:
Professor Guevara Noubir
Professor Kaushik Chowdhury