Secure edge computing vulnerabilities in smart cities sustainability using petri net and genetic algorithm-based reinforcement learning

• Published in: Intelligent systems with applications Vol. 18; p. 200216
• Main Authors: Ajao, Lukman Adewale, Apeh, Simon Tooswem
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-05-2023; Elsevier
• Subjects: Edge computing; Fog computing; Industrial internet of things; Reinforcement learning; Smart cities; Fog computing; Smart cities; Industrial internet of things; Reinforcement learning; Edge computing
• ISSN: 2667-3053; 2667-3053
• DOI: 10.1016/j.iswa.2023.200216

•Design and modeling of a secured smart cities edge computing architecture.•A secure trust-aware philosopher privacy and authentication model (STAPPA) for mitigation of the privacy breach.•Distributed authorization algorithm for the identity management in a smart city's edge computing architecture.•The free-model unsupervised reinforcement learning architecture for edge computing vulnerabilities training.•Nature-inspired algorithm (Genetic Algorithm) for the edge computing attack detection and shortest route optimization.
The Industrial Internet of Things (IIoT) revolution has emerged as a promising network that enhanced information dissemination about the city's resources. This city's resources are wirelessly connected to different constrained devices (such as sensors, robotics, and actuators). However, the communication of this wireless information is threatened by several malicious attacks, cyber-attacks, and hackers. This is due to unsecured IIoT networks that were exposed as a potential back door entry point for the attacks. Consequently, this study aims to develop a security framework for the smart cities’ sustainability edge computing vulnerabilities using Petri Net and Genetic Algorithm-Based Reinforcement Learning (GARL). First, a common trust model for addressing information outflows in the network using a distributed authorization algorithm is proposed. This algorithm is implemented on a secure framework modeling in Petri Net called secure trust-aware philosopher privacy and authentication (STAPPA) for mitigation of the privacy breach in the networks. Genetic Algorithm-based Reinforcement Learning (GARL) is used to optimize the search, detect anomalies, and shortest route during the agent learning in the environment. The detection and accuracy rate results obtained over a secure framework using reinforcement learning are 98.75, 99, 99.50, 99.75, and 100% during simulation in the network environment. The average sensitivity of the detection rate is 1.000, while the average specificity outcome is 0.868. The result of the GARL simulation model obtained shows the best distance of 238.84 * 10−3 fitness when the search space is optimized by reducing the number of chromosomes to 10 in the model. These approaches help to detect anomalies and prevent unauthorized users from accessing edge computing components in the city architecture.