Guest Editorial Special Section on Recent Advances in Security and Privacy for 6G Networks

The emergence of new disruptive technologies is paving the way towards shaping the upcoming sixth generation (6G) of wireless networks, which are envisioned to enable a large number of innovative applications over a ubiquitous, secure, unified, self-sustainable, and fully intelligent platform. These technologies include but are not limited to, virtual/augmented/mixed reality services, haptics, flying vehicles, brain-machine interface, and telepresence, to name a few. The successful operation of their associated functionalities is subject to meeting stringent network requirements, such as extremely high data rates, ultra-low latency, low complexity, uniquely small-sized designs, and high energy and spectral efficiencies. Therefore, the evolution of 6G networks will be accompanied by diverse novel technological trends, including artificial intelligence, data mining, cloud and edge computing, wireless mobile caching, network slicing, network function virtualization, as well as centralized and decentralized deep learning. While 6G wireless paradigms are envisaged to support the realization of self sustaining, self optimized networks with personalized user experience, privacy and security remain a predominant concern due to the centralized and decentralized data exchange, storage, and process, needed for the successful operation of 6G networks.


I. SECURE MASSIVE MIMO SYSTEMS
In [A1], Roth et al. designed a novel physical layer authentication scheme, that jointly incorporate the channel and process system parameters for enhanced security level. The developed framework is based on Kalman filter and follows the threshold-based principle. In particular, the Kalman filter is utilized to estimate the system and channel states, which are then leveraged as inputs to a hypothesis test for node authentication. The authors further optimized the threshold value in the hypothesis testing procedure for guaranteed security. The robustness of the authentication scheme is corroborated in several scenarios, including, a small number of antennas, massive single-input multiple-output (SIMO), and massive SIMO with channel hardening.
An efficient secure offloading mechanism was proposed by Yilmaz et al. in [A2]. In specific, they developed a cooperative mobile edge computing (MEC) scheme, which incorporates massive MIMO and non-orthogonal multiple access (NOMA) for improved communication between cell-edge users and MEC servers. The framework takes into consideration the limited computing capabilities, power budget, and security constraints at cell-edge and cell-center users, and aims to minimize the overall delay over the downlink and uplink communications.

II. PHYSICAL LAYER SECURITY IN NON-TERRESTRIAL NETWORKS
Abdrabou et al. in [A3] developed a physical layer authentication scheme for low earth orbit (LEO) satellites, through leveraging Doppler frequency shift (DS) and received power (RP) for hypothesis testing. The hypothesis testing is performed through threshold-based approach, as well as machine learning (ML) algorithms, e.g., one-class classification support vector machine. The algorithm was trained using real satellite data of legitimate nodes. The authors showed the effective impact of DS in improving the authentication rate in small elevation angles, while it was demonstrated that for large elevation angles, the RP has more effect on the authentication performance.
The performance study conducted by Erdogan et al. in [A4] aimed at quantifying the secrecy performance of different eavesdropping scenarios in an optical high altitude platform station (HAPS) setup. The authors studied the secrecy outage probability (SOP), probability of positive secrecy (PPSC), average secrecy capacity (ASC), and secrecy throughput (ST). Through the developed mathematical framework in [A4], the authors drawn useful conclusions on the design aspects to be taken into consideration in optical HAPS systems, e.g., eavesdropper's SNR, the scattered information level, zenith angle, and distance, for improved physical layer security.

III. NOVEL MACHINE LEARNING-BASED SECURITY APPROACHES
In [A5], Al-Jarrah et al. developed a new ML-based intrusion detection system (IDS) for detecting novel cyberattacks in intra-vehicle networks, specifically in controller area networks (CANs) of modern vehicles. The proposed IDS framework implements the Recurrence Plot (RP) to generate high-level representations of CAN messages incorporating both the content transported by a message and its relative context. The captured complex relationships and temporal dependencies among the messages are fed into a bespoke Neural Network, designed and trained to detect novel intrusions.
The efforts of Uysal et al. [A6] were devoted to providing a comprehensive review of the theoretical and experimental data-driven malware detection literature in the large-scale data-intensive field. The contribution of the article is two fold: (i) discussing new concepts in multi-domain to multitarget continuous learning and the challenges associated with unseen/unknown data, imbalanced data, and data scarcity; (ii) shedding light on the novel concept of explainability via visualization with a multi-labeling approach which allows identifying malware by their recipes while improving the interpretability of its decision process.
The study by Shi et al. [A7] proposed a reinforcement learning (RL)-based network slicing to maximize the total reward of accepted user requests in next generation (NextG) radio access networks (RAN). By exploiting adversarial ML, the authors introduced a novel over-the-air attack to manipulate the RL algorithm and disrupt NextG network slicing. The adversary reduces the RL algorithm's reward by observing the spectrum and building its own RL-based surrogate model to selectively jam the available resource blocks (RBs) so that the RL algorithm for the network slicing receives an incorrect reward (feedback), thereby leading to a significant performance loss of resource allocation for NextG RAN slicing. The authors designed novel defense schemes by considering various characteristics of the RL algorithms and showed the effectiveness of the designed attack and defense schemes using different benchmarks.
In the article [A8], Duong et al. presented a comprehensive overview of the state-of-the-art in quantum computing (QM). They identified several quantum-inspired ML applications for 6G networks and discussed their underlying potentials and challenges in terms of resource allocation and network security, considering their enabling technologies. This article highlighted some dominating research issues and offered future research directions for quantum-inspired ML in 6G networks. The presented study provided insights into QC with ML and offered substantial guidelines for the quantum developers and researchers of the next generation of quantum computers and how they transform the quantum ML algorithms into practical applications.

IV. SECURITY AND PRIVACY FOR SMART CITIES
The survey presented by Aldahmani et al. [A9] focused on the cyber-security of embedded IoTs in Smart Homes and highlighted relevant challenges, requirements, countermeasures, and trends. After overviewing IoT's design, objects, and standards for smart homes, the authors provided an in-depth discussion on state-of-the-art privacy and security approaches for smart homes. In addition, the article detailed the major smart home components that must be safeguarded and addressed the tiered IoTs framework and associated security concerns by introducing a taxonomy related to vulnerabilities, threats, and attacks on IoT devices in this domain. Moreover, the authors highlighted several recommended and countermeasures security solutions that can be used to keep IoT devices, networks, and applications cyber-safe and protect them against cyberattacks.
In [A10], Gheisari et al. developed a novel privacypreserving mechanism for Internet-of-Things (IoTs) devices within a smart city environment. Considering the deployment of software-defined networking (SDN) in IoT devices, the authors proposed a dynamic differential privacy scheme, which frequently selects either the Laplace distribution or the exponential distribution to protect the sensitive data produced by IoT devices. The selected privacy-preserving method of each IoT device changes every minute, ensuring sensitive data disclosure is prevented. The study demonstrated that, compared to the traditional static privacy-preserving methods, the proposed dynamic scheme is more effective in preserving network privacy and bringing flexibility to network management.

ACKNOWLEDGMENT
The Guest Editors would like to express their gratitude to all the authors for their outstanding contributions to this special section, and all the anonymous reviewers for their efforts and valuable comments, which helped to improve the quality of the articles. The Guest Editors also would like to extend their gratitude to the Editor-in-Chief, Sun Sumei, for her diligence and support throughout the entire process of this special section.