Differentially private plans enable semantic summaries of sensitive data
Differentially Private Semantic Plans for Aggregate Insight Generation
Machine Learning
Summary
It is hard to share useful summaries about sensitive data without revealing private details. The authors developed DP-SPIN, a method that creates privacy-protected summaries aligned with meaningful concepts defined outside the data. Their approach carefully adds noise to aggregate information, ensuring individuals’ privacy while allowing insights about common themes. DP-SPIN works by representing each data record as a vector of concept mentions, clipping users’ contributions, and releasing a noisy summary that can be verified publicly. They tested their method on customer reviews and complaint data, showing it provides usable insights with strong privacy guarantees.
What this means in practice
- •For customer analytics teams: Generate privacy-protected summaries of customer feedback grouped by predefined concepts across multiple datasets without exposing individual reviews.
- •For social media platform operators: Produce differentially private aggregate reports about content themes while ensuring user contribution limits and allowing public verification of outputs.
Authors
Behrooz Razeghi
Abstract
\texttt{URANIA} provides end-to-end differential privacy (DP) for summaries of data-dependent clusters. However, its cluster--keyword release does not directly provide collection-wide aggregates for semantic concepts defined independently of the protected corpus. Records may express several concepts, records expressing the same concept may be assigned to different clusters, and cluster identities need not correspond across analyses. Consequently, cluster-level statistics do not directly provide comparable measurements of predefined concepts across collections or repeated analyses. We introduce \texttt{DP-SPIN}, a trusted-curator framework for aggregate measurement and summarization over semantic concepts fixed independently of the protected target records. Each record is mapped to a bounded sparse nonnegative vector over these concepts, whose sum forms a semantic sketch. A differentially private mechanism releases a semantic plan containing admitted concepts and noisy masses; normalized semantic-support values and support bins are obtained by post-processing. For user-level privacy, each user's aggregate contribution is clipped to a fixed bound. The language model receives only the plan and fixed decoding instructions, while a public verifier checks concept mentions, reported values, comparisons, and rank claims against the released plan. The final summary is differentially private by post-processing. We establish record- and user-level DP guarantees under add/drop and replacement adjacency. We evaluate \texttt{DP-SPIN} under record-level privacy on CFPB complaint narratives, Amazon All Beauty reviews, and Yelp restaurant reviews, and under user-level privacy on Amazon and Yelp. We compare \texttt{DP-SPIN} with non-private plan and summary references, DP keyword and category histogram baselines, and a \texttt{URANIA}-style baseline with a fixed public keyword vocabulary.