Intelligent Context-aware Fog Node Discovery and Trust-based Fog Node Selection

Abstract

In today’s highly advanced technological age, edge devices are widely used. By 2030, Cisco predicts that more than 500 billion edge devices (also known in this research as fog consumers) will be in use [1]. Data from all these devices may experience significant delays when handled, processed and stored through cloud computing. To resolve this issue, fog computing is the best solution. With fog computing, processing, storage, and networking are brought to the edge of the network near fog consumers. This reduces latency, network bandwidth, and response times. Researchers have yet to address the critical challenge of identifying and selecting a reliable and relevant fog node to fog consumers. The existing approaches consider the discovery and selection of fog nodes based on the networking point of view. However, no approach addresses the use of AI-driven mechanisms for intelligent fog node discovery and selection. This research aims to propose an intelligent and distributed framework for context-aware fog node discovery and trust-based fog node selection. This research aims to discover the closest fog nodes in a context-aware manner and select a reliable fog node based on the trust value. The proposed approach is based on the distributed Fog Registry Consortium (FRC) between fog consumers and fog nodes that can facilitate the discovery and selection processes of fog nodes. To ensure that the tasks from the fog consumer are processed in a timely manner, one of the crucial aspects to consider for fog node discovery is the geographic distance between the fog node and the fog consumer as this directly impacts latency, response time, and bandwidth usage for fog consumers. Thus, location-based context awareness is one of the key decision criteria for fog node discovery to ensure that the QoS metrics are satisfied. In this research, we propose the Fog Node Discovery Engine (FNDE) within the Distributed Fog Registry (DFR), within FRC, as an intelligent and distributed fog discovery mechanism which enables a fog consumer to intelligently discover fog nodes in a context-aware manner. In this research, the KNN, K-d tree and brute force algorithms are used to discover fog nodes based on the location-based context-aware criteria of fog consumers and fog nodes. Fog node selection is a crucial aspect in the development of a fog computing system. It forms the foundation for other techniques such as resource allocation, task delegation, load balancing, and service placement. Fog consumers have the task of choosing the most suitable and reliable fog node(s) from the available options, based on specific criteria. This research presents the intelligent and reliable Fog Node Selection Engine (FNSE), which is an intelligent method to assist fog consumers to select appropriate and reliable fog nodes in a trustworthy manner. This intelligent mechanism predicts the trust value of fog nodes to help the user select a reliable fog node based on its trust value. Our selection approach is based on the trust value of the fog node based on the values of the QoS factors. If the fog node has historical information of the QoS factors provided to this fog node, then the Trust Evaluation Engine (TEE) in the FNSE is responsible to carry out the prediction of the trust value. With the trust value of fog nodes, the FNSE will be able to rank the fog node to select the most reliable fog node in the network. We propose three mechanisms: the TEE mechanism based on fuzzy logic, the TEE mechanism based on logistic regression, and the TEE mechanism based on a deep neural network. However, if the QoS values of the fog node are unknown, this means the FNSE is unable to make a meaningful selection of fog nodes. To solve the problem of the cold-start fog node, we propose the Bootstrapping Engine (BE) which is an intelligent trust-based fog node bootstrapping framework. This framework is designed to address the cold-start problem in fog computing environments which enables fog consumers to make informed and trustworthy decisions when selecting fog nodes for their applications. To address this challenge, the BE employs two key modules, namely the QoS prediction module and the reputation prediction module. The QoS prediction module utilizes the k-means clustering and KNN algorithms to predict the initial QoS values of new cold-start fog nodes. Additionally, within the reputation prediction module, we propose three AI methods to achieve the best performance and prediction results, namely fuzzy logic-based reputation prediction, regression-based reputation prediction, and deep learning-based reputation prediction to predict and evaluate the trust value of the new cold-start fog nodes. Finally, we present the simulation of the framework and the evaluation results of each proposed engine which highlight the best performance.