CAREER: Digital Encoding of Information Signals for Security with Limited Resources

Princeton University, Princeton NJ

Investigators

Abstract

Technical Description: Information signals are abundant in modern complex systems, and the encoding needed to provide security during communication and storage becomes one of many important security concerns. This research project will develop theory and technology for providing a critical degree of security to information signals, which will allow an engineer to optimize the tradeoff between the expense of security resources and the security guarantees that the system can provide. The underpinnings of this work are information theoretic in nature. The Pincipal Investigator has demonstrated the emergence of new theory for understanding partial secrecy. Much of the focus in secure communication research is on mechanisms for providing security resources, usually in the form of a private authenticated channel or an authenticated secret key. This might be accomplished through cryptography, information theoretic physical layer security, or quantum key exchange. This project will address the question of how to efficiently use those resources to protect signals (source coding). Prior information theory literature on secure source coding typically either is limited to perfect secrecy or uses normalized mutual information as a metric for the level of secrecy provided. This project has two key innovations. First, privacy guarantees come in the form of limits on what an adversary can do with eavesdropped information. This is measured, for example, by average distortion or payoff guarantees, analogous to rate-distortion theory. Second, an adversary is assumed to independently obtain additional information about the signals in real time, to assist in their attack. These innovations have already lead to interesting theoretical discoveries as well as unusual encoding and stochastic decoding mechanisms for efficiently achieving secrecy. The plan for this project involves investigating the theoretical implications of important practical constraints, such as delay in the communication, delay in the adversary's information, uncertainty about the information signal, complex signals, multi-user settings, etc. These theoretical results provide guidelines for how secrecy should be designed. To bridge theory to application, this project has two areas of emphasis. The first is a study of the use of cryptography as the security resource. This involves complexity analyses of the codecs used as well as a theoretical understanding of the consequences of differing types of resources (i.e., private channel vs. secret key). The second is the application of these technologies to secure distributed control. Broader Significance and Importance: Security is relevant to important societal infrastructure, such as energy, transportation, and communication networks. This project aims to fill an important gap in information theory and the understanding of digital communications. Although a rich mathematical theory for encoding and compression of information is known in the field, the security counterpart to this theory, which incorporates constraints on security resources, is conspicuously lacking. This project aims to provide new fundamental understanding to the field. The educational component of this project will engage in the swelling movement of high quality and freely available education resources. Princeton University has recently partnered with Coursera, a non-profit organization that collaborates with research and education institutions to provide open (free) access to high quality education. This mode of teaching not only increases outreach but enables innovative teaching features through technology. The P.I. has engaged with Coursera to create an undergraduate signals and systems course. Additionally, he will create and teach graduate courses containing research discoveries from this project. This funding will support Ph.D. students in their education and research. The P.I. has a track record of mentoring female students (two out of six current Ph.D students). Undergraduates are also incorporated into the discovery process through independent work mentoring.