Abstract
Dynamic edge-cloud repositories are core infrastructures for Internet of Things (IoT) systems, where records are continuously maintained and updated. As these repositories evolve, records may be written, modified, or deleted, while retrieval requests remain privacy sensitive. Under such dynamic conditions, rerunning the entire preprocessing procedure after every update is impractical, making incremental private information retrieval essential. However, three challenges remain: (1) existing schemes mainly target multi-server settings and lack a formal definition for the single-server case; (2) most rely on heavyweight cryptographic tools such as homomorphic encryption, limiting practical efficiency; and (3) dynamic updates often lack verifiability and forward privacy, so clients cannot verify server-side updates and later update events may leak information about past queries. To address these challenges, we propose VFIP-PIR, a verifiable and forward-private single-server incremental private information retrieval scheme for dynamic edge-cloud repositories. We first formally define encrypted single-server incremental private information retrieval for this setting, and then develop an efficient construction based on lightweight pseudorandom functions. To ensure security under dynamic updates, we introduce a dynamic Merkle tree for verifiable state evolution and combine hint refreshing with secure erasure to achieve forward privacy. Experimental results show that VFIP-PIR achieves sub-millisecond online query latency. Compared with full rebuilding, it reduces maintenance overhead by over 99%. Relative to existing incremental schemes, it reduces refresh latency by 25.4× to 45.5× and post-update online latency by 21.24× to 66.90×, while still providing verifiability and forward privacy.