Talk Title: On Using Differential Privacy

Speaker: Dr. Gabriel Kaptchuk, Assistant Professor, Computer Science Department, University of Maryland, College Park

Location: HCIL Lab (HBK 2105) and Zoom

Abstract: Differential privacy is a statistical privacy technique that injects carefully calibrated noise into datasets or computations in order to bound the maximum amount an attacker can learn about any data subjects. Since first introduction in 2006, differential privacy has rapidly become the go-to tool in statistical privacy and has been deployed across industry and government. Unfortunately, many of these deployments have been plagued by controversy and confusion—not because of technical failures, but because of the human-computer interaction challenges that theoreticians did not foresee. In this talk, we draw on a number of our recent results to illustrate some of the challenges of using differential privacy responsibly, and some promising techniques we have developed to try and overcome these challenges.

Bio: Dr. Gabriel Kaptchuk is an Assistant Professor in the Computer Science department at the University of Maryland, College Park. Dr. Kaptchuk’s work is centered in applied cryptography, but has grown to span multiple adjacent fields of inquiry, including usable security and technology policy. Dr. Kaptchuk was previously a Research Assistant Professor at Boston University and received his Ph.D. and M.S. in Computer Science from Johns Hopkins University.