学术报告|Private Set Intersection (PSI), Variants,and Applications

Private Set Intersection (PSI), Variants,
and Applications

Private Set Intersection (PSI) is a fundamental protocol in secure computation, enabling two or more parties to compute the intersection of their confidential input sets without revealing any additional information. Several PSI variants have emerged to address different privacy-preserving requirements. PSI-Payload extends the protocol by associating additional data (payloads) with intersecting elements, while PSI-Sum aggregates these payloads. PSI Cardinality focuses on computing the intersection size without revealing the elements themselves, and Unbalanced PSI optimizes efficiency for cases where one party’s set is significantly larger than the other. Private Membership Testing further refines PSI to verify whether individual elements belong to the other party’s set. To construct efficient PSI protocols and variants, we have adopted algebraic approaches and probabilistic techniques. The latter has enabled Approximate PSI (Approx-PSI), which matches “close” elements under distance metrics like Hamming or cosine similarity, significantly enhancing applicability to fuzzy data scenarios. PSI and its variants have found applications across diverse fields, including privacy-preserving business data matching, secure contact tracing, and biometric authentication. Approx-PSI, in particular, can be used in image-matching tasks, where it allows for private comparison of similar images. These advancements make PSI an indispensable tool in privacy-focused technologies.