The over-the-air (OTA) computation which utilizes the waveform-superposition property of wireless signals has been considered as a promising approach to simultaneously accomplish communication and computing tasks in multiple access channels. Equipping federated learning (FL) with the OTA computation allows distributed Artificial Intelligence of Things (AIoT) devices to collaboratively train machine learning models over wireless environment, while preserving data privacy without excessive bandwidth consumption. In fact, the appearance of jamming attacks in OTA-FL systems can severely disrupt the convergence. This paper studies the problem of jamming attacks and its countermeasures in over-the-air federated learning (OTA-FL). To this end, we propose the coordinated dropout strategy (CoDrop), which enables AIoT devices to collaboratively drop out (i.e., to refrain from transmitting) part of their gradients so that the jamming signals aggregated in received signals can be accurately measured and mitigated. Our simulation results reveal that CoDrop is effective in alleviating the negative impacts of jamming signals with significantly low dropout rates, and it is shown to converge well as compared to existing solutions under various parameter settings.

The Internet of Vehicles (IoV) enables vehicles to communicate with each other and infrastructure through basic safety messages (BSMs), thereby reducing accidents by means of early warnings and collision avoidance. In fact, IoV networks are vulnerable to positional attacks induced by malicious vehicles, which can severely compromise traffic safety and system reliability. In spite of the existing works devoted to misbehavior detection for certain types of attacks, extracting statistical features from BSMs to defend against all positional attacks in IoV networks has been largely overlooked. In this paper, we extract six types of positional attacks from the VeReMi-Extension dataset and discover hidden attack patterns. Based on these discovered patterns and our data analysis, we categorize them into the statistical features of Gaussian-like, Lévy-like sparse, and Lévy-like ordinary positional attacks. Then, we propose a three-feature misbehavior detection mechanism to distinguish between normal and malicious vehicles, where the first feature detection employs the Jarque-Bera test to identify Gaussian-like positional attacks, the second feature detection identifies Lévy-like sparse positional attacks by evaluating data sparsity, and the third feature detection uses Isolation Forest to detect Lévy-like ordinary positional attacks. Simulation results show that the proposed solution achieves promising false alarm and miss detection rates, and acts robustly across diverse parameter settings.

The over-the-air (OTA) computation which utilizes the waveform-superposition property of wireless signals has been considered as a promising approach to simultaneously accomplish communication and computing tasks in multiple access channels. Equipping federated learning (FL) with the OTA computation allows distributed Artificial Intelligence of Things (AIoT) devices to collaboratively train machine learning models over wireless environments, while preserving data privacy without excessive bandwidth consumption. However, the appearance of wardens in OTA-FL can result in model eavesdropping and the risk of information leakage. In this paper, we investigate how OTAFL can be fulfilled among AIoT devices while ensuring the covertness (i.e., the prevention of eavesdropping) against wardens. To this end, we formulate an optimization problem to maximize achievable data rates subject to covertness and transmit power constraints. Then, we first derive optimal detection thresholds to simplify the covertness constraints, and then propose the covertness-aware transmit power control (CAMEO) algorithm, which iteratively explores transmit powers that yield high target signal-to-interference-plus-noise ratios while satisfying both the covertness and transmit power constraints. Simulation results demonstrate that the CAMEO algorithm outperforms comparison schemes in terms of test accuracy, and it is also shown to converge well under various parameter settings.

Satellite networks provide higher coverage and provide ubiquitous mobile services. However, how to allocate precious satellite spectrum resources to improve better network performance has become an important issue in satellite networks. In this paper, we study the power allocation problem of the Low Earth Orbit (LEO) satellite to maximize the Supply-Demand Ratio (SDR) of the LEO satellite users while minimize the standard deviation (SD) of LEO satellite users’ SDR. We propose an Event-Driven Deep Reinforcement Learning based Power Allocation Mechanism (EDRL-PAM), which utilizes the LEO satellite’s power manager to intelligently allocate the power request for each cell of the LEO satellite. Furthermore, we propose a NOMA-based power allocation algorithm for allocating power to all LEO satellite users within the cells. The proposed EDRL-PAM fully utilizes a Deep Reinforcement Learning (DRL) technique, Deep Deterministic Policy Gradient (DDPG), to deal with stochastic arrivals of power requests in the power manager to achieve long-term optimization of the network performance. The simulation results show that our proposed EDRL-PAM can significantly improve the average data rate, and achieve the long-term optimization and fairness of network performance for satellite users.

Switching off base stations (BSs) is an effective and efficient energy-saving solution for green cellular networks. The previous works focus mainly on when to switch off BSs without sacrificing the traffic demands of current active users, and then enlarge the coverage of the stay-on cells to cover as many users as possible. Based on this objective, both constant power and transmission power of each BS become the major energy consumption sources. However, the transmission powers of enlarged cells, which have not been taken into account in previous research, are not negligible as compared to other energy consumption sources. To tackle this problem, we observe that the transmission power of one specific BS could be reduced via cooperation among two or more BSs, which is typically used to improve the throughput or enhance the spectrum efficiency in wireless systems. The challenges come mainly from how to jointly consider which BSs to switch off and how to cooperate among active-mode BSs. In this paper, we design energy-aware cooperation strategies that ensure that our system is energy-saving while satisfying user demands. To cope with sleep-mode BSs and perform cooperation among active BSs, we formulate this problem as a binary integer programming problem, and prove it is NP-hard. Based on our formulation, we derive a performance lower bound for this problem via Lagrangian Relaxation with search enumeration. Furthermore, we propose two heuristic algorithms accounting for the properties of energy savings and the constraints of bandwidth resources. The simulation results show that our algorithms outperform pure power control mechanisms that do not consider the transmission power and pure cooperation without power control in terms of the total consumed energy. We also observe that larger cooperative size does not imply a better strategy under different scenarios. Compared to the total consumed energy given that all BSs are turned on, our algorithms can save up to 60 percent of energy. This demonstrates that our methods are indeed efficient energy-saving cooperation strategies for green cellular networks.

A heterogeneous network (HetNet), which consists of macro cells and small cells, is an effective tool to improve the system spectral efficiency by exploiting the benefit of spatial reuse. However, the benefit of HetNet is limited by the asymmetry of transmit power between macro base stations (BSs) and pico BSs. Traditional solutions lead to a tradeoff between received signal strength (RSS) and load balancing. To tackle this problem, we propose a novel mechanism with joint macro BS transmit power control and load-aware user association. As such, we can realize traffic offloading from the macro cell to pico cells in a two-tier HetNet and mitigate inter-cell interference (ICI) in the network. We formulate this problem as a nonlinear integer programming problem, and prove that it is NP-hard. We then propose a polynomial-time heuristic. The simulation results show that our algorithm is indeed an efficient mechanism with better network performance in terms of average throughput and fairness.

A heterogeneous network (HetNet) with macro base stations (BSs) together with low power small cells, such as pico BSs and femto BSs, is a promising solution to enhance network capacity. In HetNet, the transmission range of pico BSs are limited due to the large transmission power difference between macro and pico BSs, and thus pico BSs are typically underutilized. Cell Range Expansion (CRE) is a way to increase the opportunity of the user equipment (UE) associations to pico BSs so that more traffic from the macro cell can be offloaded to the pico cells. However, the expansion region of pico UEs with unbalance downlink and uplink signal qualities suffer from bad downlink quality, and the macro UEs near the expansion region may also experience low uplink rates. Therefore, we propose a new scheme with rate-based CRE offsets in HetNet so that UEs can decide to their associations based on their traffic demands. We formulate this problem as a mixed integer programming problem and prove that it is NP-hard. We then propose a polynomial time heuristic algorithm to solve this problem. The simulation results show that our scheme indeed achieves good traffic offloading gain for HetNet.

IPTV multicasting is an important service for residential users. In this paper, we explore how to allocate resources to different layer-encoded IPTV programs using multicast services in relay-assisted centralized wireless access networks. We consider the scenario of two-hop relay stations, and the MAC protocol is OFDMA-based, as in IEEE 802.16m or Long Term Evolution-Advanced (LTE-A). We formulate this problem via an Integer Linear Programming model and propose a heuristic algorithm called Maximum Utility resource allocation for Relayed Multicast Services (MU-RMS) to solve this problem. The simulation results show that our algorithm performs efficiently. We also show that our performance is affected by how the forwarding strategy is performed.