After two decades of research, a certain amount of looking back at one’s career naturally happens. This retrospection has involved looking at my research from a variety of different angles. While there are many different topics I have worked on and papers I have written over the years, I have gravitated towards two different views of my research… let us call them the traditional and the philosophical.
The traditional cut is your typical engineering faculty member’s approach: many different projects, funded by a variety of different agencies, but nonetheless coalescing into a handful of major research themes. For those wanting a traditional view of my research, please take a look at my project page.
The philosophical cut is my newfound way of looking at my career: looking across all of my work, what are the hidden, deeper threads or insights that recur across these projects? After some introspection and a journey up the mountain, I have found that my research has several recurring themes:
Know your enemy and know yourself…
An essential part of every security researcher’s repertoire should be putting themselves in the role of the adversary. By understanding the motivation and tools that an opponent comes to the table with, one is equipped to devise better tactics (i.e. tools) and strategies for dealing with the problems an opponent may present. Much of my research has involved adversarial modeling in one form or another, and from those models we have devised techniques to protect systems or even devise ways to fight back against the enemy. One of my favorite examples of how we achieved this was a project I affectionately call the “four ounce network” (named after the Tai Chi Chuan principle “A strength of four ounces can defeat a force of a thousand pounds”), in which we realized a method to create a wireless communication network in spite of the presence of a significantly more powerful interfering opponent. This network exists as a timing-channel overlay on top of a normal wireless communication network and was designed so that if the opponent does not attack, the system communicates as normal, but if it does interfere, then the lack of communication itself serves as a means to communicate. As another example, we have shown that it is essential, in many security problems, to use decoys (or honeypots) to force the opponent to reveal their presence and their strategies, which then make it possible to apply to appropriate countermeasures.
If you build it…
Well, “they” might not come, but chances are that you will learn something in the process that was not obvious, and that by itself opens up many opportunities for further research. I consider myself a theory-meets-systems researcher, and constantly I am amazed at how many new ideas come to the surface when we build a system or conduct an experiment. My work on physical layer security was originally conceived by playing around with acoustic transmitters and a sound card, and this eventually led to us building one of the first prototypes of physical layer security using the channel estimation carried out in 802.11. In the process of building our real system, and subsequent work on Mach-Zehnder interferometry for optical security, I have identified many new directions for exploration, and am currently working to tackle the simple and fundamental question “Is it better to design a system on a theoretically solid physical foundation that is difficult to build, or a system based on a slightly weaker physical foundation yet can be built more easily?” This question lies at the heart of both wireless and optical physical layer security. Our work on building sensor systems uncovered key challenges related to energy utilization, which led us to develop new communication and computing architectures that can minimize energy usage for many emerging Internet of Things applications. Whether it is working on integrating ultra-sound sensors with cars to detect empty parking spots or building cognitive radios capable of scanning vast amounts of spectrum, I always look for the new research problems that come from building a real system, and which can spur future theory and experimental research.
When life gives you lemons, make lemonade…
The world is replete with things that we don’t want, hurdles that we think we must overcome in order to move forward. But, if one looks back at the history of science and engineering, there are many examples where a slight change in perspective turned a challenge into an opportunity. This has been particularly true with wireless communications, such as the use of Rake processing (which turns multipath into a diversity-enhancing benefit) or MIMO antennas (which turn scatter-induced Rayleigh fading into a capacity benefit). In our work on physical layer security, we recognized that the rich multipath environment typical of wireless communication scenarios leads to a frequency-selective, dispersive and location-specific source of common randomness from which one can build new authentication and confidentiality security services. Or, as another example, in our work on secrecy and privacy, we often use noise as a means to cover or obscure signals that we want to be hidden from an adversary. Throughout much of my work involving game-theoretic modeling, we frequently find that, rather than sticking to clearly defined strategies, the introduction of randomness in agent action (mixed strategies) leads to “stability” that is not possible when agents do not act randomly (pure strategies). Understanding how one can nudge a system to desirable behaviors by taking advantage of concepts like noise and randomness is at the heart of much of my work.
Everything is connected to everything else…
This basic ecological principle reveals itself to us in science and engineering in so many ways. In communications, we can improve our data rates by increasing the amount of resources we use (e.g. transmit power or bandwidth), but that means we affect others (causing increased interference or taking bandwidth from others). In networks and systems, we find that entities are connected to each other and often one agent’s actions will depend upon another agent being able to complete their action. The science challenge I am tackling is to find and understand these connections—a problem that is often very difficult in its own right as many times these connections are not easily probed. The engineering challenge is to then use what we know about the science behind these connections to control or tear down these connections as we see need in order to achieve a desired objective. Recently, we have investigated the resilience of communication networks using a combination of graph-theoretic and game-theoretic tools, which have led to the implementation of distributed power control algorithms that make peer to peer wireless networks more resilient to being punctured by interference or protocol attacks. Other work has applied similar graph-theoretic tools to arrive at non-invasive classification tools that can be used to diagnose patients with Parkinson’s disease. Other work that I have been involved in has explored the use of bargaining and fairness in the design of systems with multiple objectives (such as radar and communication objectives), where addressing one objective impacts the system’s ability to address another objective.