Introduction to the Special Section on Resiliency for AI-enabled Smart Critical Infrastructures for 5G and Beyond

5G positioning makes use of a number of edge devices to analyze the signals from user equipment in order to locate it accurately. Due to the sensitivity of position and privacy requirements of involved parties in location-based services (LBSs), new challenges are raised for preserving privacy with multiple objectives, not only for users, but also for LBS providers. However, the literature still lacks studies on multi-objective privacy preservation in 5G. For protecting privacy for both 5G positioning and LBS provision in an integrated way, the article “Privacy Protection in 5G Positioning and Location-Based Services Based on SGX” by Yan et al. proposed a novel light-weight scheme by employing the Intel Software Guard Extensions (SGX) to secure position calculation in a fusion center and preserve data privacy of multiple LBS providers without frequent key exchange through secret sharing. Serious security analysis and performance evaluation over a real-world database show its sound security properties and high efficiency.

Real-time smart grid monitoring is essential to enhance the resiliency and operational efficiency of power equipment and critical modern infrastructures. The article by Li et al., titled “Resource Orchestration of Cloud-edge based Smart Grid Fault Detection,” provided an AI-enabled real-time smart grid detection utilizing edge/cloud computing and cellular networks. In particular, the authors designed multiple lightweight neural networks to provide flexible detection services and consequently, conceive an optimal communication and computational resource allocation method with the constraints of data transmission and processing latency.

With the development of AI-enabled smart critical Infrastructures for 5G and B5G, the infrastructure will no longer simply forward data. Still, it will also assume the function of dynamic routing of data. The article by Zhang et al., titled “Toward Data Transmission Security based on Proxy Broadcast Re-encryption in Edge Collaboration,” addressed the security issue of dynamic routing in the infrastructure, and proposed an end-to-end data security transmission method based on proxy re-encryption and broadcast encryption. This work also enhanced the functionality, security, and performance of the proposed scheme with online/offline techniques as well as trusted execution environment techniques.

Orienting the high resiliency and availability requirements of 5G network infrastructure, the article by Yi et al., titled “Resilient Deployment of Smart Nodes for Improving Confident Information Coverage in 5G IoT,” focused on the network resilience deployment problem. To obtain the optimal deployment, the mixed-integer linear programming models (CICILP-COST) and (CICILP-ERROR) were proposed based on the CIC model. The proposed models were solved by the variable relaxation algorithm (CICVR-COST) and dichotomous search algorithm (CICDS-ERROR), respectively. Simulations showed that the proposed models can obtain a lower cost-optimal deployment.

In the article titled “Intelligent Video Ingestion for Real-time Traffic Monitoring,” Zhang et al. proposed a light-weight Deep Reinforcement Learning (DRL) based approach for the cameras deployed at strategic places and prime junctions in an intelligent transportation system (ITS), which can enhance the quality of the video ingestion process of a traffic monitoring system by adjusting the video bitrate adaptively in a real-time manner. Distinguished from the existing bitrate adjusting approaches, the proposed approach can overcome the bias incurred by deterministic discretization of candidate bitrates by adjusting the video bitrate with more fine-grained control from a continuous action space, thus significantly improving the Quality-of-Service (QoS).

In the article titled “Digital Twin-Enabled AI Enhancement in Smart Critical Infrastructures for 5G,” Gai et al. addressed the task of optimizing the AI training accuracy to satisfy the needs of digital-twin, when training time and energy consumption are limited. One of the main contributions of the article was proposing a new strategy called EAHAS (Energy-aware High Accuracy Strategy), which takes energy-saving, efficiency, and accuracy into consideration at the same time. The authors evaluated the strategy under various network parameter settings, and the results showed that training accuracy was enhanced by 12% compared with traditional mechanisms.

The work presented in “Sleeping Cell Detection for Resiliency Enhancements in 5G/B5G Mobile Edge-Cloud Computing Networks” by Ming et al., contributed to the building of a mobile edge-cloud computing system to dynamically detect the sleeping cells, to improve the security and ease the resiliency degradation concerns for smart critical infrastructure in 5G/B5G networks. The authors tackled the challenging issue of complex hardware and software failures which are difficult to detect, by proposing a semi-supervised learning-based framework. The new framework can improve both the recovery proportion and recovery speed, to enhance the resiliency of small base stations. Evaluation of the proposed framework was carried out by comparing it with other existing sleeping cell detection schemes with real-world datasets, and the results proved that the resiliency of SBSs was improved.

In the article titled “Improving Quality of Service in 5G Resilient Communication with cellular structure of Smartphones,” Sangaiah et al. described the development of the optimal working mode in 5G resilient communication with the cellular structure of smart phones. D2D communication modes are used to improve the throughput of the network. The authors formulated the problem as finding the optimal working modes under various configurations through extensive simulations. The proposed method was successfully evaluated under different system conditions and parameters, and the results showed that its efficiency outperforms the QoS-D2D method.

To meet the high anonymity when exchanging data through 5G and WiFi technologies, the article “A Secure and Anonymous Communicate Scheme over the Internet of Things” by Sun et al. proposed a peer-to-peer network model for secure communications. The main contributions include: (1) using a typical crowds system to meet IoT user demands under constrained resources; (2) designing a lightweight communication scheme using two kinds of virtual space; and (3) carrying out extensive experiments over a typical open source-based prototype implementation. Experimental results showed that the proposed mechanism can meet the expected requirements.

Narrowband IoT (NB-IoT) cellular is an important technology for 5G/B5G, however, its coverage and connectivity is limited in dense urban areas. The article “NB-IoT Coverage and Sensor Node Connectivity in Dense Urban Environments: An Empirical Study” by Yau et al. presented their empirical study that focuses on evaluating the performance of NB-IoT in high-rising apartment building environments in Hong Kong. By installing over one hundred sensors that utilize a commercial NB-IoT network to uploading data, the authors collected and analyzed the signal data. The measurements revealed the correlation between connectivity and environments, e.g., NB-IoT signal strength is improved in denser environments. Another finding is that inter-cell interference may degrade the connectivity of NB-IoT.