A Privacy-Preserving Data Sharing Framework Driven by Blockchain for Smart Childcare Subsidy Systems

Abstract

With the continuing development of digital government and public childcare-assistance programs, cross-departmental subsidy services must jointly address sensitive-data minimization, dynamic authorization, fraudulent-application identification, and accountable auditing. This paper proposes a blockchain-driven privacy-preserving data-sharing framework that combines a consortium blockchain, on-chain evidence with off-chain ciphertext storage, zero-knowledge eligibility proofs, attribute-based encryption, proxy re-encryption, secure aggregation, federated learning, and smart-contract auditing. The contribution is a scenario-specific system integration and governance design rather than a new cryptographic primitive. In addition to the original Hyperledger Fabric feasibility test, the revision adds literature-grounded validation against peer-reviewed studies of childcare-subsidy administration and administrative-data evaluation, component implementation-status disclosure, a 2048-bit proof-of-knowledge precomputation microbenchmark, non-IID federated-learning tests, and colluding-node poisoning tests. The original prototype reaches 718 TPS and 236 ms average latency at 400 concurrent requests, 91.8% federated-learning accuracy, and 138 ms on-chain index retrieval for 20,000 audit records. In the added five-run tests, strong non-IID data retain an F1 score of 82.69% +/- 0.09%; adaptive clipping and robust aggregation retain F1 scores of 84.04% and 84.14% under one and two colluding sign-flip clients, respectively. Offline commitment precomputation reduces the online arithmetic time for four eligibility subproofs from 8.03 +/- 0.34 ms to 0.06 +/- 0.01 ms. These results support the feasibility of the integrated framework while clarifying that production deployment still requires authorized local data, audited cryptographic libraries, and multi-orderer field validation.