Part I: Wiretap Channels with Random States Part II: Differential Privacy as a Mutual Information Constraint

In the first part, we propose a new encoding scheme for the wiretap channel setting where a random state is known non-causally to the encoder but not necessarily to the decoder (the Gelfand-Pinsker setting). This setting combines two scenarios with celebrated results in information theory–the wiretap channel and the Gelfand-Pinsker channel–and it so happens that essentially the same encoding scheme is optimal for both scenarios individually. A decade ago, Chen and Han Vinck analyzed the secrecy rates achieved by that same encoding scheme. Quite surprisingly, Chia and El Gamal more recently showed that a different encoding scheme which explicitly extracts a key from the state to be used for encryption can sometimes outperform that scheme. In this work, we show a simple superposition code design which performs at least as well as both of these competing schemes, accomplishing the key agreement concept in a more subtle way through the selection of the auxiliary random variables.

In the second part, an information theoretic notion of differential privacy will be presented, which establishes an equivalence between differential privacy and a mutual information constraint. This intuitive definition in terms of mutual information admits very simple proofs of several important properties of differential privacy.
Paul Cuff received the B.S. degree in electrical engineering from Brigham Young University, Provo, UT, in 2004 and the M.S. and Ph.D. degrees in electrical engineering from Stanford University in 2006 and 2009. His Ph.D. research advisor was Thomas Cover. Since 2009 he has been an Assistant Professor of Electrical Engineering at Princeton University.

Over the years Dr. Cuff has interacted with industry in both the technology and the financial sectors, spending summers at Google, Microsoft Research, and elsewhere, and giving talks at a number of hedge funds. In 2005, while in graduate school, he co-founded a tech startup called Adaptive Hearing Solutions with Bernard Widrow centered around signal processing technology. This venture began with the winning of the Stanford business plan competition.

As a graduate student, Dr. Cuff was awarded the ISIT 2008 Student Paper Award for his work titled "communication Requirements for Generating Correlated Random Variables." This work has led to fruitful and unexpected avenues of research in secure source coding. As faculty, he received the NSF Career Award in 2014 and the AFOSR Young Investigator Program Award in 2015.