Deep learning Models for Trust management in Social Internet of Things

Abstract

Listen

Translate article

Introduction: The growing prevalence of IoT has resulted in both human-to-thing(H2T) communication and thing-to-thing(T2T) communication. In recent years, a new paradigm merging IoT with social networks has arisen, termed the Social Internet of Things (SIoT), where devices are not only intelligent but also socially conscious. Trust is an important element in SIoT and it allows reliable automatic communication between items and trust management is also an essential aspect for secure communications between IoT devices, In the dynamic SIoT environment, a traditional trust framework's static or heuristic approach is ineffective. Objectives: To study the Deep Learning (DL) Models for trust management in SIoT and to suggest a hybrid model to predict the trustworthiness of the IoT devices in SIoT Networks. Methods: A combined strategy is suggested in this study using Graph Neural Networks (GNN) and Recurrent Neural networks (RNN) to predict the trustworthiness of SIoT using vast high-dimensional data and detect specific trends in social communications. Results: The finding shows that the hybrid GNN and RNN model outperforms with an accuracy of 91.8% over the standalone GNN and RNN methods. Conclusions: The integrated model that includes both GNN and RNN techniques gives better results in terms of accuracy, precision, recall, and F1-score over the other methods assessed separately. The ability of deep learning to interpret complex multidimensional data and adapt to the highly dynamic nature of social and physical interaction is significant in maintaining trust within SIoT networks.