A NOVEL ALGORITHMIC FRAMEWORK FOR SECURING IOT TRUST MANAGEMENT AGAINST EMERGING TRUST-RELATED ATTACKS

Abstract

The Internet of Things (IoT) is spreading very quickly and has connected and automated a huge range of areas, from transportation and healthcare to industrial systems.  IoT networks, on the other hand, are open to a variety of trust-based threats that can slow down systems, stop people from talking to each other, and make security less reliable because they are spread out and different.  Attacks like blackhole, grayhole, floods, and TDMA-based outages change trust measures and packet-forwarding behaviours, which has a direct effect on trust management systems.  To make IoT trust management safer from these new dangers, this study suggests a new algorithmic framework that combines advanced feature selection and mixed classification methods to improve the accuracy of detection. The first step in the method is to use the WSN-DS dataset, which has examples of attacks on wireless sensor networks that are related to trust and attacks that are not related to trust.  As part of data preparation, columns are dropped, labels are encoded, and one-hot encoding is done. Next, Min-Max normalization is used to make sure that the size of all the features is the same.  Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) are used alone and together to reduce the number of dimensions while keeping their prediction power.  Trust3Net is a mixed PSO+GA method that improves model generalization by choosing the best feature groups. Several classification methods are looked at in the study. These are Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), Neural Networks, Long Short-Term Memory (LSTM), and Convolutional Neural Networks (CNN).  Two suggested deep learning-based designs are created: Trust4Net and Trust3Net. Trust3Net uses a dual feature selection approach to improve its performance.  Comparative research shows that combining metaheuristic feature selection with deep learning makes IoT trust management systems much more resistant to new threats.  The suggested framework provides a strong defence system that can be expanded and changed to fit real-world IoT operations where maintaining trust is important for maintaining operating efficiency and security.