Trusted Devices Analysis and Classification Model for Social Internet of Things

Advanced technology utilization in the heterogeneous device communication in Social Internet of Things (SIoT) field is emerging research area in order to provide efficient communication based on social relationship among the objects in a trusted way. In the field of SIoT in order to establish trusted connection and communication by mitigating from the untrusted activities involved and device participation requires an efficient framework and real time machine leaning model equipped into device intelligence in order classify and also to provide the secure and trust management relationship among the heterogonous interconnected sensor devices. An intelligent solution is needed in an adaptive SIoT environment to identify and classify the malignant and untrusted nodes. Hence efficient trust-based model is required where the trustable nodes are identified using an Intelligent Trust Management Learning Model (ITMLM) for the Social Internet of things is propsed called ITMLM-SIoT. The proposed system integrates numerous factors of trust classification such as direct and indirect trust, transaction factors and other social modeling trust factors. The implementation uses both the human intelligence and device artificial intelligence for trust management classification. The proposed system uses a combination of dynamic (interaction-based) and static (graph-based) dataset collection mechanism with reference to object and device communication in an trusted manner that helps in limiting resource overheads of a adaptive and dynamic approach for effective trust model classification by benefiting from its higher accuracy compared to a other existing approach. The paper provides analysis of the trust management model classification and shows the effectiveness of the proposed model in handling different untrust scenarios while requiring limited storage and computational resources in SIoT devices. The proposed machine learning models compares the efficient performance. The proposed model achieved 88.87% and 89.76% of accuracy for the given data set.