CSE Colloquium: A System-Level Framework for Privacy

Abstract: Privacy in the digital age has become increasingly difficult to achieve as technologies that capitalize on facial recognition, location services, and personal health tracking become mainstream. While there is consensus on the importance of building privacy into systems that deal with sensitive information, our ability to reason about system-level privacy is severely limited. In this talk, I introduce wringing, a new computer architecture approach to building privacy in systems that minimizes information leakage while still maintaining utility of the privatized data. Using this framework, I demonstrate how wringing can preserve privacy of program traces. Next, I detail how reverse engineering attacks can compromise privacy in localization pipelines and how wringing can be used to safeguard privacy in these situations. Finally, I touch upon future work with themes including performant privacy codesign, private machine perception, and verifiably private architectures. 

Biography: Deeksha Dangwal is a Ph.D. candidate at UC Santa Barbara in the Department of Computer Science. She studies computer architecture and is interested in the design of private systems and applications. Her interdisciplinary research spans computer architecture, privacy, computer vision, and machine learning. She is a Rising Star in EECS 2020 and her work has been published in ASPLOS, IEEE Micro Top Picks, ISCA, PLDI, and FPL.