A Trust Prediction Mechanism in Edge Communications using Optimized Support Vector Regression

Abstract

The number of edge devices is increasing every day in the fog computing environment. According to Gartner's prediction, around 42 billion edge devices will be involved in digital communications by 2025. Different kinds of edge devices will be involved in various applications such as healthcare, transportation, and education to provide services at anytime and anywhere to the user. At the same time, attackers are trying to intrude into the communication system by taking the advantage of heterogeneity of devices. Consequently, trust management among edge devices is one of the major security concerns in identifying untrustworthy activities in the communication system. This paper proposes a mechanism to predict the trust values of every edge device participating in the communication based on the attributes using support vector regression (SVR). Accuracy, loss rate, recall, precision, and F-measure are used to assess the performance of the suggested model on various data samples of various sizes. Performance comparisons with existing machine learning models demonstrate superior results with various iteration counts. The proposed model attained 99.98% accuracy, 0.0048 loss rate, 99.96% precision, 100% recall, 99.96% F-Measure and took almost 356 seconds for 100 iterations.