Quantifying privacy risk: clean maths and messy reality

Speaker: Prof Anand D Sarwate, Rutgers University
Series: C.U. Ethics in Mathematics Society (CUEiMS)

"Differential privacy” (DP) has emerged as a de facto standard for assessing and controlling privacy risk in information processing systems. At its heart, DP is based on hypothesis testing, which is a basic topic in statistics. In this talk, I will explain a bit of how DP works and the cost of privacy (the privacy-utility tradeoff). I will then talk about how challenges that arise when we try to use differential privacy in practice: gaps between theory and reality, the complexity of engineering systems and software, “privacy washing,” and more.