Federated Learning With Small and Large Models With Privacy-Preserving Data Space for Holographic Internet of Things in Consumer Electronics

Holographic Internet of Things (IoT) aggregates virtual and augmented reality to provide real-time modeling that improves the user experience of consumer electronic products and applications. The incorporated technologies support third-party applications for which heterogeneous privacy-preserving features are required. Considering this factor, a Modeling Space Privacy (MSP) is introduced in this article using ternary homomorphic encryption (THE). The proposed privacy scheme encourages space and component privacy using independent hashes using HE. Privacy is retained using the ternary operation between the components and space to ensure maximum security of IoT model representations. Third-party applications, components, and spaces need to follow a unison between the IoT platforms to improve security. The components/space elements are discarded from the holographic representations to prevent anonymous access/views to the actual models. The federated learning process in the proposed scheme differs from the ternary process. Therefore, the proposed scheme is reliable in employing conditional HE to improve the privacy of holographic IoT platforms irrespective of multi-party inclusions. From the comparative values, the proposed scheme is optimal in reducing the latency by 13.412% and privacy process complexity by 12.664% for the maximum devices considered.