Guest Editorial Special Issue on Smart IoT System: Opportunities by Linking Cloud, Edge, and AI
IEEE Internet of Things Journal
R ECENTLY, the Internet of Things (IoT) technologies have made their entrances into many fields, such as smart city, healthcare, intelligent transportation, forest protection, and environmental monitoring. However, the current IoT system is facing great difficulty to efficiently handle the huge data generated from IoT devices. It has become challenging to ensure low latency, energy efficiency, and so on. To cope with the huge data and reply promptly and accurately, a recent trend is to deploy
... lltrained artificial intelligence (AI) model on edge servers, whose capacities are somewhat bigger than those of IoT devices. However, since the training of AI demands huge computation and memory resources, the training process is preferred to be done in the cloud. This link among cloud, edge, and AI poses many challenges that call for novel approaches and rethinking of the entire architecture, communication, and processing to meet requirements in latency, reliability, and so on. This special issue aims to create a platform for researchers from both academia and industry to disseminate state-of-theart results and to advance the use of cloud, edge, and AI to build intelligent IoT systems. The response to our call for this special issue was overwhelming, as we received in total 164 submissions from around the world. During the review process, each article was assigned to and reviewed by at least three experts in the field, with a rigorous multiround review process. Thanks to the great support from the former Editor-in-Chief, Prof. Xuemin (Sherman) Shen, and the current Editor-in-Chief, Prof. Honggang Wang, and the dedicated work of numerous reviewers, we were able to accept 29 excellent articles covering various topics in IoT-enabled CAVs. In the following, we will introduce these articles and highlight their main contributions. In the article "A provenance-aware distributed trust model for resilient unmanned aerial vehicle networks," Ge et al. study the security issues and trust assessment for unmanned aerial vehicle networks and propose a distributed trust model based on a provenance-aware approach. The proposed model collects the observational evidence for the distribute trust evaluation and leverages the source of the packet to identify the malicious nodes behaves, such as black hole, packet injection attack, and modification attack. In the article "D2D-enabled mobile-edge computation offloading for multiuser IoT network," Yang et al. design Digital Object Identifier 10.1109/JIOT.2021 computational offloading schemes in D2D networks. The designed scheme considers collaboration constraints among users and dynamic resource availability to fully utilize the available resources of idle devices at the edge. In the article "FraudTrip: Taxi fraudulent trip detection from corresponding trajectories," Ding et al. propose a system, called FraudTrip, which detects "unmetered" taxi trips based on a novel fraud detection algorithm and a maximum fraudulent trajectory construction algorithm, without the help of taximeters. In the article "A cluster-based multidimensional approach for detecting attacks on connected vehicles," D'Angelo et al. provide two algorithms that implement a data-driven anomaly detection system. The first algorithm (cluster-based learning algorithm) is used to learn the behavior of messages passing on the CAN bus, for baselining purposes, while the second one (data-driven anomaly detection algorithm) is used to perform real-time classification of such messages (licit or illicit) for early alerting in the presence of malicious usages. In the article "Efficient and privacy-preserving decision tree classification for health monitoring systems," Liang et al. propose an efficient and privacy-preserving decision tree classification scheme (PPDT) for health monitoring systems. The privacy-preserving decision tree classification is achieved by searching the encrypted indexes with encrypted biomedical data. In the article "A new subspace clustering strategy for AI-based data analysis in IoT system," Cui et al. propose a post-process strategy of subspace clustering for taking account of sparsity and connectivity. The proposed strategy defines close neighbors as having more common neighbors and higher coefficients neighbors, where the close neighbors are selected according to the nondominated sorting algorithm, and prunes the intersubspace connections by eliminating incorrect or useless connections. In the article "Automatic detection of congestive heart failure based on a hybrid deep learning algorithm in the Internet of Medical Things," Ning et al. propose an automatic CHF detection model based on a hybrid deep learning algorithm which is composed of convolutional neural network (CNN) and recursive neural network (RNN). The proposed model classifies normal sinus heart rate signals and CHF signals based on electrocardiograph (ECG) and time-frequency spectra during the RR interval. The accuracy of this algorithm is 99.93%, sensitivity is 99.85%, and specificity is 100% when analyzing 5-min ECG signals.