Edge Intelligence for the Industrial Internet of Things

8 T Industrial Internet of Things (IIoT) is deemed to be the revolutionary progress in the new Industry 4.0 era. With the development of machine sensors, computer systems, and modern telematics, IIoT provides comprehensive insight into wide productivity areas including manufacturing, energy management, and other industrial processes, which can reduce costs, improve efficiency, and ultimately gain business profits. The growth of the industrial machines and network scale has brought excessive burden on the automation and intellectualization of the industrial process, and consequently voluminous quantities of big data. The traditional cloud model provides advanced analytics performed in powerful central servers by gathering massive data and sending the processed results back. Although the model has proved to be effective and successful for legacy applications, it leaves a dilemma of the communication latency in the cloudhost services for IIoT applications, which often require prompt response and mobility support. The data traffic that has to be routed to remote cloud servers via uncontrolled network paths also increases the risks of compromised data confidentiality and privacy leakage. Edge computing provides a solution to the dilemma by providing an infrastructure close to the network edge between the IIoT devices and cloud servers that can provide similar services as the cloud model, such as auxiliary computing and storage capacity. The edge differs from the cloud by reducing the latency as it is much closer to the data source, and thus can improve the quality of service (QoS) for IIoT applications. The latest innovation in artificial intelligence, blockchain and other big data techniques further envision the future IIoT to create inherent intelligence that machines can work and adapt to the environment without human intervention. To upgrade the current IIoT ecosystem with these kinds of new capabilities (e.g., self-monitoring, autonomous operation), this Special Issue (SI) proposes to present and highlight the advances and the latest intelligent technologies, implementations, and applications in the field of edge-based IIoT, so as to move the theoretical and practical frontiers forward for a deeper understanding from both the academic and industrial viewpoints. In response to the Call for Papers, 14 outstanding articles have been selected for this SI after a careful review process. The articles in this SI are classified into three categories: • Security and privacy in edge intelligent IIoT • Analytics architectures, frameworks, and models for intelligent edge-based IIoT • Machine-learning-based optimization methods for edge intelligent IIoT In the first article, “Edge Intelligence and Blockchain Empowered 5G Beyond for Industrial Internet of Things,” the authors create an intelligent and secure edge intelligence framework for fifth generation (5G) beyond, which is further applied in IIoT. The study mainly realizes the collaboration of edge resources in a novel cross-domain sharing approach. Moreover, deep learning methods and algorithms are also exploited in the proposed edge intelligence framework to help system managers recognize industrial environments, understand application requirements, and optimize resource scheduling strategies. To ensure secure edge services, blockchain is utilized with a newly devised credit-differentiated transaction approval mechanism, which significantly improves the efficiency and flexibility of the blockchain consensus process. This article is a pioneering study in leveraging edge intelligence and blockchain for 5G beyond wireless communications (also called 6G). The proposed resource sharing mechanism is truly superior in tackling the resource constraints in terms of types, geographical distributions, and service capacities. The study will have high impact on the technology advancement of 5G beyond and 6G wireless communications. In the second article, “Toward Data Security in Edge Intelligent IIoT,” the authors aim at data security in IIoT and analyze four main challenges in this scenario, including reliable storage, convenient usage, efficient search, and secure deletion. They then address the issues by presenting corresponding solutions to each challenge. An optimized framework, facilitating the underlying IIoT in data acquisition, processing, and transition, is proposed, which enjoys stronger security and better efficiency. The authors also implement the instantiations to each possible solution on both a laptop and a Raspberry Pi to show the efficiency. In the third article, “Privacy-Preserving Image Retrieval for Medical IoT Systems: A Blockchain-Based Approach,” the authors focus on the limitations of existing studies on retrieval of medical data, which either fail to protect sensitive information of medical images or are limited to a single image data provider. In this article, the authors propose a blockchain-based system for medical image retrieval with privacy protection. Using the emerging blockchain techniques, they present the layered architecture and threat model of the proposed system. In order to accommodate large-size images with storage-constrained blocks, the authors capture a carefully selected feature vector to represent a medical image. They also design a customized transaction structure to record feature vectors on blockchain, which protects the privacy of both medical images and image features. The IIoT ecosystem can efficiently optimize several relevant computational factors such as execution time, price or economic cost, and energy consumption in edge computing to improve productivity, efficiency, and reliability. EdgE IntEllIgEncE for thE IndustrIal IntErnEt of thIngs GUEST EDITORIAL