Locally private frequency estimation of physical symptoms for infectious disease analysis in Internet of Medical Things.

Frequency estimation of physical symptoms for peoples is the most direct way to analyze and predict infectious diseases. In Internet of medical Things (IoMT), it is efficient and convenient for users to report their physical symptoms to hospitals or disease prevention departments by various mobile devices. Unfortunately, it usually brings leakage risk of these symptoms since data receivers may be untrusted. As a strong metric for health privacy, local differential privacy (LDP) requires that users should perturb their symptoms to prevent the risk. However, the widely-used data structure called sketch for frequency estimation does not satisfy the specified requirement. In this paper, we firstly define the problem of frequency estimation of physical symptoms under LDP. Then, we propose four different protocols, i.e., CMS-LDP, FCS-LDP, CS-LDP and FAS-LDP to solve the above problem. Next, we demonstrate that the designed protocols satisfy LDP and unbiased estimation. We also present two approaches to implement the key component (i.e., universal hash functions) of protocols. Finally, we conduct experiments to evaluate four protocols on two real-world datasets, representing two different distributions of physical symptoms. The results show that CMS-LDP and CS-LDP have relatively optimal utility for frequency estimation of physical symptoms in IoMT.

An Exception Handling Approach for Privacy-Preserving Service Recommendation Failure in a Cloud Environment.

Service recommendation has become an effective way to quickly extract insightful information from massive data. However, in the cloud environment, the quality of service (QoS) data used to make recommendation decisions are often monitored by distributed sensors and stored in different cloud platforms. In this situation, integrating these distributed data (monitored by remote sensors) across different platforms while guaranteeing user privacy is an important but challenging task, for the successful service recommendation in the cloud environment. Locality-Sensitive Hashing (LSH) is a promising way to achieve the abovementioned data integration and privacy-preservation goals, while current LSH-based recommendation studies seldom consider the possible recommendation failures and hence reduce the robustness of recommender systems significantly. In view of this challenge, we develop a new LSH variant, named converse LSH, and then suggest an exception handling approach for recommendation failures based on the converse LSH technique. Finally, we conduct several simulated experiments based on the well-known dataset, i.e., Movielens to prove the effectiveness and efficiency of our approach.