Abstract:

Based on the current development status of privacy computing both domestically and internationally, this paper summarizes the research progress in the field of privacy computing interconnectivity.Utilizing a systemic architecture perspective, it elaborates on the implementation pathways of interconnectivity technology across three levels: the＆quot;application layer, protocol layer, and communication layer＆quot;.Considering the complexities arising from the current characteristics of interconnectivity platforms, including intricate computational principles and diverse architectures, an innovative Adaptation mechanism is proposed as an interconnectivity framework.Through the design and implementation of key technologies, this framework not only ensures the realization of existing functionalities but also addresses the compatibility issues posed by different architectures.By means of experiments conducted in various scenarios encompassing traditional machine learning, horizontal federated learning, and vertical federated learning, and considering dimensions such as data volume and feature distribution, the effectiveness and viability of the trustworthy interconnectivity gateway under the Adaptation framework have been demonstrated.

Key words: privacy-preserving computation, interconnection, system architecture, federated learning, trusted gateway