Universal Shuffle Asymptotics, Part III: Dominant-Block Quotient Geometry and Hybrid Gaussian--Compound-Poisson Limits in Finite-Alphabet Shuffle Privacy
Alex Shvets

TL;DR
This paper completes the finite-alphabet shuffle privacy theory by analyzing dominant-block geometry, revealing Gaussian and compound-Poisson limits, and establishing sharp rates and boundary behaviors in neighboring shuffle experiments.
Contribution
It introduces the dominant-block quotient geometry framework, extending the limit theory to arbitrary block sizes and overlaps, and characterizes the hybrid Gaussian--compound-Poisson limits in shuffle privacy.
Findings
Decomposition of the limiting experiment into Gaussian and compound-Poisson components.
Identification of regimes where the quotient description fully determines the privacy-curve.
Proof that the O(n^{-1/2}) rate is sharp, with conditions for O(n^{-1}) rate restoration.
Abstract
Part I of this series (arXiv:2602.09029) establishes a sharp Gaussian (LAN/GDP) limit theory for neighboring shuffle experiments in the fixed full-support regime. Part II (arXiv:2603.10073) identifies the first universality-breaking frontier: critical Poisson, Skellam, and multivariate compound-Poisson regimes. The present paper completes the finite-alphabet weak-limit theory by identifying the dominant-block quotient geometry that governs neighboring shuffle experiments. We treat dominant blocks of arbitrary finite size, allow overlap between the dominant output sets under the two neighboring hypotheses, and show that the limiting experiment decomposes according to this geometry: projecting onto the sum of the dominant tangent spaces yields a Gaussian factor, while quotienting by those same tangent spaces isolates a compound-Poisson jump field in the rare block. We also identify the…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsRandom Matrices and Applications · Stochastic processes and statistical mechanics · Probability and Risk Models
