An efficient Trust Related Attack Detection Model based on Machine Learning for Social Internet of Things

Abstract

The Social Internet of Things (SIoT) can be defined as an IoT where objects can establish social links with each other, in which the capability of humans and objects to select, determine, and use services in the IoT is improved. Hence, without a security mechanism to ensure trustworthy SIoT nodes' interactions, this paradigm cannot reach enough popularity to be a well-established technology. A Trust Management (TM) model becomes a major challenge in SIoT systems to ensure qualified and secure services. Nevertheless, the defined TM models don't show their performance towards trust related attacks performed at the aim of disrupting the management of the trust values. In this work, we propose a TM model dedicated not only to identify trustworthy nodes, but also to detect and prevent malicious attacks. Based on Machine Learning (ML) techniques, this model can identify these attacks by learning proposed trust features that are derived from the description of malicious node behaviors. Experimentation results generated with simulated data sets based on real data concur the effectiveness of our proposed model.