JISE

With the technologies of wireless mobile networks and wireless devices, users have faced a severe issue of data overload in application usage. Preference recommendation of applications is viewed as an effective method to solve such issues, which mainly relies on users’ interests in an application. However, preference recommendation of applications ignores what functions users like to utilize. This paper aims to address the data overload issue by analyzing users’ preference of functional classes for devices in wireless mobile networks and proposes a privacy scheme of block-stream service for DEX files based on federated learning. We first propose a formalized model of application functions to provide insight into applications. We calculate correlation weights of functions in mobile applications by using users’ operation behavior data. The homomorphic encryption technologies are also utilized to prevent privacy leakage of model parameters in this paper. Finally, models involved in the proposed scheme are stored in the blockchain to support trusted storage and traceability services. Experiment results are presented to verify the effectiveness of our proposed scheme. Compared with other block-streaming loading schemes, the proposed scheme in this paper has a pretty good performance in saving storage space.