Owing to the easy deployment and mobile flexibility, Unmanned Aerial Vehicle (UAV) assisted Mobile Edge Computing (MEC) has been deemed as one potential technology for handling the computation-intensive tasks at terminal devices (TDs). In this work, a MEC architecture assisted by UAVs is designed which achieves efficient offloading, computing, and downloading for tasks from multiple TDs via aerial to aerial collaboration of two UAVs. In this architecture, the task offloading process contains two parts, i.e., the offloading from TDs to a mobile UAV which flies around TDs, and the offloading from the mobile UAV to a hovering UAV which hovers in the air. The computing tasks from TDs will be divided into three parts allocated to the TDs themselves, and both two UAVs. Upon completion of computation, the computation results are downloaded to the TDs. The optimization objective is to seek for an optimal task division strategy to attain the weighted total energy consumption minimization for all devices. Since the formulated optimization problem is not convex, we develop a two-step iteration algorithm which jointly optimizes computing frequency, task allocation volume, as well as UAV's trajectory based on the method of block coordinate descent. Simulation results confirm the effectiveness and performance advantages of the designed algorithm.

In Internet of Things(IoT) systems, ensuring the secure exchange of information between devices from different domains is crucial. Current cross-domain authentication schemes based on a single blockchain struggle to meet the confidentiality requirements for data and information exchange in large-scale IoT systems. This paper proposes a blockchain-based identity authentication scheme (BCPPAS) for IoT, featuring dual-chain collaboration, and designs a novel certificateless aggregate signature algorithm to address complex certificate management and key escrow issues. The edge server is capable of aggregating different signatures to achieve batch authentication, markedly improving authentication efficiency and reducing computational and storage overhead. Also, BCPPAS is designed to avoid costly bilinear pairing operations, providing less computational overhead for IoT devices with limited resources. To protect the privacy of IoT devices, BCPPAS uses the pseudonym instead of the real identity. Finally, efficiency of BCPPAS are demonstrated through theoretical analysis and experiments.

The Internet of Things (IoT) connects numerous de-vices and sensors, generating data with significant informational and economic value. However, data silos hinder effective data utilization and trading, leading to the dispersion of data across various devices and systems. Additionally, traditional third-party trading models face challenges related to data security and trust. To address these issues, this paper proposes a secure data trading framework based on a consortium blockchain and designs a corresponding solution. Specifically, it introduces the integration of zero-knowledge proofs into the smart contract scheme for authenticity and integrity verification of transaction data. From the perspective of IoT device users, this paper aims to enable secure data trading through a decentralized platform, using off-chain storage methods to reduce the blockchain's data burden while ensuring security and privacy. Off-chain storage encrypts and securely stores sensitive data, recording only necessary information on the blockchain, effectively protecting user privacy. To validate the practicality of the proposed solution, experiments were conducted using Hyperledger Fabric, demonstrating its feasibility in facilitating secure storage and trustworthy trading of IoT data. Finally, this study analyzes the experimental results and offers valuable insights for future research.

The awareness of data privacy preservation in the Internet of Things (IoT) environment and the amount of IoT data production, are growing almost in parallel with each other. As a privacy-preserving framework, Federated Learning (FL) allows many participants to collaboratively build machine learning models while ensuring that their raw data remains local and undisclosed. However, as the devices charged in data collection are deployed in different IoT environments, we also face a significant challenge i.e., dealing with non-independently and identically distributed (non-IID) data. If the data is not distributed uniformly among the participants, it may lead to a significant performance degradation of the generated global model, which is far from the case when the data is distributed uniformly. To address this challenge, this study innovatively designs the Enhancing Federated Learning algorithm with Client-Specific Data Augmentation (FedSDA). The FedSDA matches clients by servers, and clients train local models using augmented datasets to overcome the negative influence mainly caused by non-IID, which consequently enhances the model accuracy. Our simulation experiments on the datasets Fashion-MNIST and CIFAR-10 ultimately demonstrate that FedSDA outperforms contemporary state-of-the-art FL strategies with similar design characteristics.

With the growth of the Internet of Things (IoT), millions of users, devices, and applications compose a complex and heterogeneous network, which increases the complexity of digital identity management. Traditional centralized digital identity management systems (DIMS) confront single points of failure and privacy leakages. The emergence of blockchain technology presents an opportunity for DIMS to handle the single point of failure problem associated with centralized architectures. However, the transparency inherent in blockchain technology still exposes DIMS to privacy leakages. In this paper, we propose the privacy-protected IoT DIMS (PPID), a novel blockchain-based distributed identity system to protect the privacy of on-chain identity data. The PPID achieves the unlinkability of identity-credential-verification. Specifically, the PPID adopts the Zero Knowledge Proof (ZKP) algorithm and Shamir secret sharing (SSS) to safeguard privacy security, resist replay attacks, and ensure data integrity. Finally, we evaluate the performance of ZKP computation in PPID, as well as the transaction fees of smart contract on the Ethereum blockchain.

Accurate and robust sea clutter modeling and parameter estimation are foundational for target detection. Traditional modeling methods rely on measured data, while clutter modeling based on radar settings and oceanographic observations is an alternative. This letter leverages high-resolution sea clutter data from the Sea-Detecting Radar Data-Sharing Program (SDRDSP) to address this challenge. Three distribution types, which are generalized Pareto distribution (GPD), K distribution, and compound-Gaussian model with inverse Gaussian (CGIG), are considered. Using random forest (RF), we identify the most discriminative factors for distribution type classification: range and azimuth resolution cell (RARC), grazing angle, wave speed, wind speed, and significant wave height (SWH). Building on this, a stacking ensemble learning framework is proposed to effectively regress the shape and scale parameters from these optimized input features. Experiments validate the effectiveness of the proposed approach in distribution type classification and parameter estimation.

Adaptive broadband digital beamforming (ABDBF) is an essential topic in the realm of array antenna for radar systems, because the array antenna with ABDBF could obtain wide swath and high azimuth resolution. Performance degradation at low snapshots is a serious problem for ABDBF. In this letter, the broadband beamforming weight generation network (BWGN) is proposed to quickly generate weights for ABDBF under low signal snapshot scenarios. The BWGN leverages the complex convolutional neural network to represent the mapping between input signals and output weights, which avoids the operation of covariance matrix and its inversion, thus speeding up the generation of weights. Compared with the existing neural network-based broadband beamforming method, wideband beamforming prediction network (WBPNet), training BWGN saves 71.45% of the time. Furthermore, progressive learning is introduced to the training process of BWGN, namely PL-BWGN, which further reduces training time of BWGN to 0.7529 h (h: hours). Simulation results demonstrate performance superiority of the proposed method compared with existing beamformers under low snapshot scenarios.

Sea-surface small target detection is challenging for the reason that the small target with low signal-to-clutter ratio (SCR) would be submerged by strong sea spikes in the complex marine environment. Against this background, a two-stage detection method using multi-feature-fusion-based binary classifier and SCR enhancement is proposed in this paper. In stage 1, the binary classifier extracts multi features in the temporal, frequency and time-frequency domains using convolutional neural networks (CNNs) and performs weighted feature fusion by minimizing cross entropy loss. Afterwards, in stage 2, SCR enhancement is divided into two steps, which are clutter suppression by modified orthogonal projection (MOP) and inter-frame incoherent accumulation by Kalman filter, respectively. Specifically, according to the range cells classified as clutter in stage 1, MOP is performed in the range-Doppler (RD) domain to improve SCR by using the projection operator composed by main clutter components via first-order difference method. Then, Kalman filter combines the predict state after RD compensation and observation state with the gain coefficient to realize target energy accumulation and SCR stabilization. Finally, the final decision is made by the constant false alarm rate (CFAR) detector. The simulation results indicate that the proposed binary classifier attains a significant binary classification performance improvement, laying foundations for effective sea clutter suppression and SCR enhancement so as to improve the final detection performance of target location and Doppler frequency.

Surface electromyography (sEMG) signal is a physiological electrical signal produced by muscle contraction. Different gestures can be effectively recognized from the characteristics of the sEMG signal. Currently, convolutional neural networks (CNNs) have been widely used in sEMG gesture recognition systems due to their capabilities in acquiring spatial features of sEMG signals. However, these classical CNNs are inefficient in extracting temporal correlation that resides in the time serials of sEMG signals, which is definitely important for gesture recognition. To overcome such a drawback of traditional CNN-based gesture recognition methods, we propose a multichannel hybrid deep learning model for gesture recognition by combining the multichannel CNNs with a gated recurrent unit (GRU). Specifically, we use multiple CNNs to preprocess the original multichannel EMG signals in a one-by-one manner to obtain the spatial features in the current observing window. The outputs of the multiple CNNs are concatenated and fed to a temporal-feature extracting module, which is designed by cascading a GRU with an attention mechanism. Through the GRU, the temporal features of successive signal frames can be established, while the attention mechanism is introduced to further focus on the key information in recognizing the gestures, which is beneficial to improve the robustness and accuracy of the model. Experiments show that the recognition accuracy of the proposed method reaches 97.6% and 96.7% on the Ninapro DB2 and Ninapro DB5 datasets, respectively. Compared with the classical CNN method, the performance improvement is 2.9% and xx% higher than that of the traditona CNN model, respectively.

Abstract The sum‐difference coarray (SDCA) is the union of the sum coarray (SCA) and difference coarray (DCA), which has higher degrees‐of‐freedom (DOF) than that of the DCA, resulting in a better direction‐of‐arrival (DOA) estimation performance. However, existing passive sparse arrays require spatial and temporal information to construct SDCA. In this study, a mirrored coprime array (MCA) is designed to implement SDCA using only spatial information. First, the SCA and DCA are recovered from the vectorised covariance matrix via the transform matrix. A Tikhonov regularisation method is proposed to reduce the rank‐deficiency effect of the transform matrix. The SCA has the potential to fill the holes in the DCA by adjusting the mirror position since the mirror determines the virtual sensor locations of the SCA. Then, the closed‐form expressions of the mirror position and virtual array aperture are derived for the hole‐free SDCA. The consecutive lags of the optimised SDCA are much larger than those of the DCA, significantly increasing the DOF. Numerical simulations verify that the MCA outperforms the non‐mirrored one with respect to the DOA estimation accuracy and resolution.

Multi-hop transmissions are desirable in realizing large-scale long-distance covert wireless communications, since a single-hop transmission cannot fully satisfy the covertness requirement even with high transmit power. Against this background, this work compares amplify-and-forward (AF) and decode-and-forward (DF) relaying strategies by examining their achievable effective throughput taking into the covertness quality-of-service. To this end, we first present a framework of maximizing the effective throughput with the assumption that each relay adopts equal transmit power to seek mathematical tractability. With the number of relays and each relay’s transmit power optimized, our results reveal that DF relaying outperforms AF relaying in the considered multi-hop covert communications, in terms of achieving a higher effective throughput with a smaller optimal number of relays. This is mainly due to that regardless of the same detection performance at the warden Willie for AF and DF relaying under the same condition, AF relays amplify both information and noise signals, while DF relays only forward the information signals. In addition, our examinations show that DF relaying with independent codewords achieves a higher effective throughput with a fewer number of relays relative to the DF relaying with a single-codeword. Furthermore, we find that the optimal number of relays increases as the desired covert communication distance increases or the covertness constraint becomes stringent.

Remote sensing image scene classification is to label remote sensing images as a specific scene category by understanding the semantic information of the images. It is an essential link in remote sensing image analysis and interpretation and has important research value. Convolutional neural networks (CNNs) have been dominant in remote sensing image scene classification due to their powerful feature extraction capabilities. The general trend has been to make deeper and wider CNN architectures to achieve higher classification accuracy. However, these advances to improve accuracy enlarge the network, creating too many parameters and high computational costs. Large models are difficult to deploy on resource-constrained edge devices for practical applications. Furthermore, CNNs can effectively capture local information but are weak in extracting global features. To overcome these drawbacks, we propose a novel knowledge distillation (KD) based method by employing Swin Transformer as a teacher network for guiding MobileNetV2 with Parameter-Free Attention (MobileNetV2-PFA). First, we modify MobileNetV2 by introducing PFA into the inverted bottleneck block; this improvement helps the model learn more latent and robust features without extra parameters. Second, Swin Transformer is an excellent architecture for capturing long-range dependencies via shifted window-based attention. So, we utilize the long-range dependency information from the Swin Transformer to assist MobileNetV2-PFA training through KD. Experimental results on the challenging NWPU-RESISC45 dataset show that the proposed method outperforms the original MobileNetV2 in classification accuracy with low computational consumption.

Abstract Remote sensing scene classification aims to automatically assign a specific semantic label to each image. It is challenging to classify remote sensing scene images due to the images' diversity and rich spatial information. Recently, convolutional neural networks have been widely used to overcome these difficulties, such as the famous Visual Geometry Group (VGG) network. However, the VGG network with local receptive fields cannot model the global information of remote sensing images well. It also needs a large number of parameters and floating point operations to achieve satisfactory accuracy. To overcome these challenges, we introduce the self‐attention mechanism to the VGG network. Specifically, we replace the last four convolutional layers in the VGG‐19 network with two cascaded self‐attention blocks, each consisting of two multi‐head self‐attention (MHSA) layers with the residual network structure. The new structure can simultaneously explore the local and global information from remote sensing scenes. Such improvements not only reduce model parameters but also improve the classification performance. The effectiveness of the proposed method is validated through experiments on four public data sets, i.e., NaSC‐TG2, WHU‐RS19, AID and EuroSAT.

In this study, a novel multi‐beam pattern synthesis algorithm is proposed. The algorithm is divided into three steps. First, the pattern synthesis is performed for each beam without element excitation amplitude constraints, and form a element excitations amplitude matrix (EEAM). Second, the kernel principal component analysis (KPCA) technique is used to acquire a group of common element excitation amplitudes (CEEA). Finally, the semi‐definite relaxation technique is employed to obtain the element excitation phase of each beam. The KPCA is a kind of principal component extraction method. Compared with the traditional method, the kernel function and the kernel parameter selection criterion used in the proposed study are designed to ensure that the extracted principal component can hold more than 80% of the information of the EEAM, which means that the acquired CEEA can bestly characterise the EEAM, hence resulting in a better synthesised pattern. In addition, the use of KPCA is a quasi‐analytical process, which also speeds up the overall algorithm. Compared to the iterative multi‐beam pattern synthesis algorithm, nearly half of the synthesis time is reduced. Through several sets of synthesised examples, and compared with some classical algorithms, this algorithm proves its superiority in the comprehensive effect and calculation time.

Two fast image super-resolution reconstruction algorithms are proposed based on multigrid (MG) and Krylov subspace accelerative algorithms. After briefly introduction of image super-resolution reconstruction model, MG and Krylov subspace accelerative algorithms, two accelerative MG algorithm named MG-CG and MG-GMRES are proposed to solve the sparse symmetric positive definite and non-symmetric linear equation, which are often occurred in image super-resolution reconstruction. The restriction, prolongation and smoothing operators of each algorithm are thoroughly studied and the convergence is analyzed respectively. Experimental results demonstrate that the proposed algorithms can greatly improve the convergence rate compared with MG, two Krylov accelerative algorithms and Richardson iteration.