A Flexible and Verifiable Keyword PIR Scheme for Cloud–Edge–Terminal Collaboration in AIoT

As a cloud storage side in the cloud–edge–terminal collaboration, which empowers the Artificial Intelligence of Things (AIoT), the accuracy of data retrieval and data privacy in the cloud can significantly impact the quality of service in AIoT. Typically, the cloud facilitates data sharing through keyword-based private information retrieval (PIR). However, these keywords may contain the privacy of patients, causing the server to gain patients’ privacy during database retrieval, resulting in privacy exposure. Concurrently, malicious users seek to access more data sets than those corresponding to the keywords. It is worth to consider the construction of a secure and private retrieval system in AIoT. To protect the privacy of AIoT, this article proposes two multikeyword PIR schemes: 1) the fuzzy multikeyword PIR scheme and 2) the fine-grained flexible multikeyword PIR (FGF-mkPIR) scheme. The fuzzy multikeyword PIR scheme utilizes the proposed batch oblivious pseudo-random function (B-OPRF) based on OTE
$^{n}_{1}$
to implement the batch search. If one of the
$n$
keywords in a data set matches a requested keyword, the corresponding data sets are returned to the user, achieving fuzzy retrieval. The FGF-mkPIR scheme incorporates the proposed batch flexible OPRF (BF-OPRF) algorithm, wherein
$k$
out of the
$n$
keywords in the data set must match the
$k$
requested keywords from users for the corresponding data sets to be returned to the user. Additionally, the cloud server may tamper with the data, and the correctness of the data is periodically verified using a verifiable mechanism. The effectiveness and performance of the proposed schemes are validated through experiments and theoretical analysis.