In this paper, deceptive signal-assisted private split learning is investigated. In our model, several edge devices jointly perform collaborative training, and some eavesdroppers aim to collect the model and data information from devices. To prevent the eavesdroppers from collecting model and data information, a subset of devices can transmit deceptive signals. Therefore, it is necessary to determine the subset of devices used for deceptive signal transmission, the subset of model training devices, and the models assigned to each model training device. This problem is formulated as an optimization problem whose goal is to minimize the information leaked to eavesdroppers while meeting the model training energy consumption and delay constraints. To solve this problem, we propose a soft actor-critic deep reinforcement learning framework with intrinsic curiosity module and cross-attention (ICM-CA) that enables a centralized agent to determine the model training devices, the deceptive signal transmission devices, the transmit power, and sub-models assigned to each model training device without knowing the position and monitoring probability of eavesdroppers. The proposed method uses an ICM module to encourage the server to explore novel actions and states and a CA module to determine the importance of each historical state-action pair thus improving training efficiency. Simulation results demonstrate that the proposed method improves the convergence rate by up to 3× and reduces the information leaked to eavesdroppers by up to 13% compared to the traditional SAC algorithm.

In this paper, a novel three dimensional (3D) positioning framework of fluid antenna system (FAS)-enabled unmanned aerial vehicles (UAVs) is developed. In the proposed framework, a set of controlled UAVs including an active UAV and four FAS-enabled passive UAVs cooperatively estimate the real-time 3D position of a target UAV. Here, the active UAV transmits a measurement signal to the passive UAVs via the reflection from the target UAV. Each passive UAV estimates the distance of the active-target-passive UAV link and selects an antenna port to share the distance information with the base station (BS), which calculates the real-time position of the target UAV. As the target UAV is moving due to its task operation, the controlled UAVs must optimize their trajectories and select optimal antenna port for transmitting the positioning information, aiming to estimate the real-time position of the target UAV. We formulate this problem as an optimization problem whose goal is to minimize the target UAV positioning error via optimizing the trajectories of all controlled UAVs and antenna port selection of passive UAVs. To address this problem, an attention-based recurrent multi-agent reinforcement learning (AR-MARL) scheme is proposed, which enables each controlled UAV to use the local Q function to determine its trajectory and antenna port while optimizing the target UAV positioning performance without knowing the trajectories and antenna port selections of other controlled UAVs. Different from current MARL methods that use feedforward neural networks to approximate Q functions, the proposed method uses a recurrent neural network (RNN) that incorporates historical state-action pairs of each controlled UAV, and an attention mechanism to analyze the importance of these historical state-action pairs, thus improving the global Q function approximation accuracy and the target UAV positioning accuracy. Simulation results show that the proposed scheme can reduce the average positioning error by up to 17.5% and 58.5% compared to the value decomposition based-MARL scheme with FAS and the proposed AR-MARL method without FAS.

In this paper, a secure and communication-efficient clustered federated learning (CFL) design is proposed. In our model, several base stations (BSs) with heterogeneous task-handling capabilities and multiple users with non-independent and identically distributed (non-IID) data jointly perform CFL training incorporating differential privacy (DP) techniques. Since each BS can process only a subset of the learning tasks and has limited wireless resource blocks (RBs) to allocate to users for federated learning (FL) model parameter transmission, it is necessary to jointly optimize RB allocation and user scheduling for CFL performance optimization. Meanwhile, our considered CFL method requires devices to use their limited data and FL model information to determine their task identities, which may introduce additional communication overhead. We formulate an optimization problem whose goal is to minimize the training loss of all learning tasks while considering device clustering, RB allocation, DP noise, and FL model transmission delay. To solve the problem, we propose a novel dynamic penalty function assisted value decomposed multi-agent reinforcement learning (DPVD-MARL) algorithm that enables distributed BSs to independently determine their connected users, RBs, and DP noise of the connected users but jointly minimize the training loss of all learning tasks across all BSs. Different from the existing MARL methods that assign a large penalty for infeasible actions, we propose a novel penalty assignment scheme that assigns penalty depending on the number of devices that cannot meet communication constraints (e.g., delay), which can guide the MARL scheme to quickly find valid actions, thus improving the convergence speed. Simulation results show that the DPVD-MARL can improve the convergence rate by up to 20% and the ultimate accumulated rewards by 15% compared to independent Q-learning.

In this paper, a novel resilient unmanned aerial vehicle (UAV) framework that enables UAVs to efficiently adjust their trajectories and antenna ports to serve disconnected users due to unexpected accidents is designed. In the proposed framework, a set of UAVs equipped with fluid antennas provide service for ground users. At the beginning, each UAV optimizes its three dimensional (3D) location and selects an antenna port to maximize the sum data rate of all users. During the service period, several UAVs may not be able to continue to serve ground users due to unexpected accidents. The remaining UAVs must adjust their trajectories and antenna ports to provide communication services for the users originally served by UAVs with accidents. This problem is formulated as an optimization problem that aims to maximize the total data rates of all users during the entire service period including the period that all UAVs can provide service, the period that some UAVs cannot provide service and the remaining UAVs must adjust their trajectories and antenna ports, and the period that the remaining UAVs find fixed locations to serve users. To solve this problem, an attention and gate recurrent unit (GRU) based reinforcement learning (AGRL) method is designed. In this method, the GRUs are utilized to capture previous UAV actions including trajectories and antenna port selections and states. The transformer is used to analyze the importance of previous UAV actions and states, thus further improving the total data rates of all users. To further improve the training speed of the designed AGRL method, we mathematically derive the optimally initial UAV locations. Simulation results show that the proposed AGRL method can improve the expected data rate of all users by up to 9.89% and 10.19% compared to the value function decomposition RL (VDRL) method and the proposed AGRL method without optimizing antenna port selection.

In this paper, the framework of fluid antenna system (FAS)-assisted three dimensional (3D) passive unmanned aerial vehicle (UAV) positioning is developed. In the proposed framework, a set of controlled UAVs including an active UAV and four FAS-assisted passive UAVs, as well as a ground base station (BS) cooperatively estimate the real-time 3D position of a target UAV. Here, the active UAV transmits a measurement signal to the passive UAVs. This signal is reflected via the target UAV and received by the passive UAVs. Each passive UAV estimates the distance of the active-target-passive UAV link and selects an antenna port to share the distance information with the BS. The BS calculates the real-time position of the target UAV. As the target UAV is moving due to its task operation, the controlled UAVs must optimize their trajectories and select optimal antenna port for transmitting the positioning information, aiming to estimate the real-time position of the target UAV. We formulate an optimization problem that optimizes the trajectories of all controlled UAVs and antenna port selection of passive UAVs with the aim of minimizing the target UAV positioning error. To address this problem, an attention-based recurrent multiagent reinforcement learning (AR-MARL) scheme is proposed. In the proposed method, a recurrent neural network (RNN) acts as a local Q function of each controlled UAV to capture its historical state-action pairs, and a transformer is used to analyze the importance of these historical state-action pairs, thus improving the global$\mathbf{Q}$function approximation accuracy, thereby further improving the positioning accuracy. Simulation results show that the proposed AR-MARL scheme can reduce the average positioning error by up to 17.5 % and 58.5 % compared to the VD-MARL scheme and the proposed method without FAS.