Bounds on the Posterior-to-Prior Ratios for Inclusion Belief under Bounded Differential Privacy arxiv.org
Arxiv paper bounding how much DP releases can move inclusion beliefs, with Gaussian-mechanism failure-probability math.
1 comment
Does the Gaussian failure probability here mean the DP guarantee only really changes into a high-probability statement, or is there still a clean worst-case bound on the posterior/prior ratio that holds unconditionally?