PolySE: Efficient Fuzzy Searchable Encryption with Pattern Hidden for Cloud-IoT

The Internet of Things (IoT) deployments generate continuous streams of device logs and sensor readings that are offloaded to the cloud. Devices are resource-constrained and clouds are only partially trusted, so data are encrypted before upload. Operators must still search these ciphertexts - and queries are often approximate. Searchable Encryption (SE) allows users to securely outsource their data to an untrusted cloud server without revealing sensitive information. However, most existing SE protocols only support exact single-keyword search, which significantly limits their practical applicability. Moreover, most SE protocols inevitably leak access or search patterns to the server. Recent studies show that such leakage can be exploited to infer sensitive information about the data user. To address these issues, this paper introduces PolySE, a fuzzy searchable encryption (FSE) scheme that allows users to securely perform fuzzy search over encrypted data. PolySE uses locality-sensitive hashing (AccuracyLSH), simple hashing, and polynomial encoding in the preprocessing phase to ensure accuracy and efficiency. Afterwards, the data user and the cloud server run a secure oblivious polynomial evaluation (OPE) protocol to obtain the search result. The security analysis shows that PolySE is secure against semi-honest adversaries and minimizes information leakage to the cloud server. Overall, PolySE targets IoT data pipelines: it supports accurate fuzzy queries over encrypted device logs with low client-side cost, constant-size per-query communication, and practical pattern hiding, making it deployable on resource-constrained gateways. Finally, we compare PolySE with state-of-the-art schemes to demonstrate its improvements in query accuracy, search time, and storage overhead. For a dataset with 35,000 keyword-document pairs, PolySE improves query accuracy by 6.5% and 12.5%, reduces query time by 85% and 86%, and decreases storage overhead by 86% and 80% compared to PIPEs and EliMFS-E, respectively.