Radial Basis Function Neural Networks Towards Electronic Trust Quantification

Trust is a broad term applicable in various contexts. Trust between electronic entities is complex to quantify, particularly in intricate networks. Traditional trust algorithms rely on historical trust values, requiring storage and access to past transaction details. Contextual variations further complicate trust calculation. Challenges in calculating trust include heightened computational complexities, managing information storage, and securing access to extensive datasets crucial for evaluation. This study will explore the implementation of a Radial Basis Function Neural Network (RBFNNet) to evaluate trust. This neural network effectively approximates functions, addressing complex computational challenges. Its efficacy depends on ample training data for modeling trust values. Synthetic data generation becomes crucial to surmount the scarcity of trust datasets. This study introduces a new definition of trust between electronic entities, a seven-step Pure Synthetic Trust Data Set Generation (PSTDG) framework for generating artificial trust datasets, a new definition for validating generated trust data, and a method with three steps to authenticate the model for generating data. A process consisting of four steps was developed to design an RBFNNet model for determining trust values between electronic entities. Two experiments were conducted to determine trust values using the RBFNNet. A trust dataset was generated using the PSTDG-PeerTrust model, which incorporates the fundamental principles of trust assessment outlined in the PeerTrust model. The validation process was completed, and a new RBFNNet model PeerTrustRBFNNet was developed, utilizing optimal hyperparameters. During the second phase of experimentation, the Amazon Relational Database Service was used to exhibit the efficiency of the proposed approach in tackling real-world problems. The study determined that the PSTDG framework-generated models create valid synthetic trust datasets, and an RBFNNet effectively computes trust in digital environments. Moreover, novel definitions of trust and synthetic trust dataset validation were developed, contributing to the advancement of trust assessment methodologies in various contexts.