Sandaru Jayawardana, Ming Ding, Kanchana Thilakarathna · ACM SIGSAC Conference on Computer and Communications Security (CCS) 2026 · 2026
DOI: 10.1145/3830454.3846420
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Repeated collection of evolving numerical data can reveal additional information over time due to temporal correlation. We propose a deviation-aware stateful mechanism that coordinates consecutive releases using Local Differential Privacy randomizers. The mechanism probabilistically reuses the previous privatized output based on the overlap between consecutive reporting distributions, and preserves each report's baseline marginal distribution. We evaluate utility using mean squared error (MSE) and incremental disclosure using conditional mutual information. Preliminary results on electricity meter data show that increasing coordination reduces incremental disclosure while maintaining comparable per-report MSE. At moderate-to-high coordination levels, the mechanism achieves lower incremental disclosure than independent perturbation and permanent memoization.
No comments yet — start the discussion below.