Overview
Explore a thought-provoking conference talk by Marika Swanberg from Boston University on the intersection of data privacy, machine learning, and legal rights. Delve into the concept of the "Right to Be Forgotten" and its implications for data control and confidentiality. Examine the two facets of data erasure, the challenges of machine unlearning, and the speaker's contributions to the definitional landscape. Gain insights into deletion-as-confidentiality, adaptive history independence, and the transition from adaptive PP to deletion. This talk, part of the Workshop on Differential Privacy and Statistical Data Analysis, offers a comprehensive look at the complex issues surrounding data privacy in the digital age.
Syllabus
Intro
Legal Right to Erasure
Two Facets of the Right to Erasure
Machine Unlearning
Related Work
Our Contributions: Definitional Landscape
Setup
Deletion-as-Confidentiality
Intuition: History Independence
DP Statistics Adaptive History Independence Claim. This DP controller is not adaptively history independent
Our Definition
From Adaptive PP to Deletion
Taught by
Fields Institute