Verifiable and robust query processing for multi-owner secret-shared databases under malicious owners

Abstract Privacy-preserving set computation enables collaborative analytics over distributed datasets without disclosing sensitive information. Recent systems such as PRISM support efficient private set intersection and aggregation over outsourced secret-shared databases. However, they implicitly rely on the assumption that all data owners are honest and do not deviate from the protocol. This assumption is often unrealistic in multi-owner settings, where a small fraction of malicious owners may distribute inconsistent shares, outsource data that violate predefined attribute-level constraints, issue unauthorized queries, or interfere with aggregation-result confirmation, thereby compromising the correctness of query outcomes. To address these challenges, we propose a security-enhanced query framework that integrates verifiable data outsourcing with query-level governance for multi-owner secret-shared databases. The framework preserves information-theoretic privacy while integrating verifiable secret sharing and zero-knowledge proofs to ensure the consistency and attribute-level input validity of outsourced data shares. In addition, under an honest-majority assumption, we introduce a lightweight query-scoped consensus mechanism to collectively confirm query admissibility and aggregation-result consistency. By incorporating behavior-aware coordinator selection, the mechanism improves robustness against abnormal or malicious participants while avoiding heavyweight Byzantine fault-tolerant protocols. Experimental results show that, on workloads involving up to 50 data owners and datasets of up to 5 million records, the proposed framework incurs only moderate overhead compared with PRISM, while substantially improving robustness in the presence of malicious participants.