Federated Unlearning (FUL) focuses on client data and computing power to offer a privacy-preserving solution. However, high computational demands, complex incentive mechanisms, and disparities in client-side computing power often lead to long times and higher costs. To address these challenges, many existing methods rely on server-side knowledge distillation that solely removes the updates of the target client, overlooking the privacy embedded in the contributions of other clients, which can lead to privacy leakage. In this work, we introduce DPUL, a novel server-side unlearning method that deeply unlearns all influential weights to prevent privacy pitfalls. Our approach comprises three components: (i) identifying high-weight parameters by filtering client update magnitudes, and rolling them back to ensure deep removal. (ii) leveraging the variational autoencoder (VAE) to reconstruct and eliminate low-weight parameters. (iii) utilizing a projection-based technique to recover the model. Experimental results on four datasets demonstrate that DPUL surpasses state-of-the-art baselines, providing a 1%-5% improvement in accuracy and up to 12x reduction in time cost.

In response to sudden high-traffic signal demand caused by massive device access in urban Internet of Things environments, Unmanned Aerial Base Stations (UAV-BSs), as dynamic network nodes, can effectively enhance the coverage capacity and quality of communication network services. However, how to efficiently deploy UAV-BSs in complex urban environments while meeting the differentiated communication needs of common and special areas remains an urgent challenge. In this paper, we propose an Average Hamming Distance Modified Particle Swarm Optimization (AHDPSO) algorithm based on the similarity calculation of DNA sequences, which firstly matrices the position and velocity information, and then calculates the average Hamming distance between particles using DNA mapping sequences to identify ’outlier points’. Further, the search guidance coefficientc3is introduced to quantify the guiding effect of ’outlier points’ on the global search, and the values ofc1,c2, andc3are dynamically adjusted by combining with the chaotic mapping, so as to balance the exploratory and developmental capabilities of the algorithm. Compared with the original particle swarm optimization algorithm, matrix particle swarm optimization algorithm, and seven other improved particle swarm optimization algorithms, the experimental results show that AHDPSO can quickly converge to the optimal solution. Compared with the traditional PSO algorithm, the absolute improvement values in the coverage of the entire region and special regions are 7.82% and 7.29%, respectively. It also shows good stability in different scenarios, indicating that the proposed algorithm has significant advantages in convergence speed, coverage, and stability.

State estimation for nonlinear systems using Finite Impulse Response (FIR) filters presents two key challenges: constructing effective filter structures and deriving optimal solutions. This paper introduces a novel nonlinear FIR filtering approach that addresses these challenges through the Koopman operator framework. The Koopman operator, a prominent data-driven method, provides intrinsic coordinates that enable global linearization of nonlinear dynamics. This transformation yields a linearized system that accurately represents the original non-linear behavior, offering new possibilities for nonlinear FIR filter design. To enhance robustness, we model the linearization error as bounded noise, enabling the filter to accommodate significant modeling mismatches within defined limits. Combined with the system’s Gaussian noise, this hybrid uncertainty description leverages the FIR structure’s inherent high degree of freedom. The optimal filter gain is then determined through multi-objective optimization. Simulation results demonstrate the proposed filter’s exceptional robustness against both modeling uncertainties and noise characterization errors, confirming its practical applicability.

Reconfigurable intelligent surfaces (RIS) are emerging as a promising technology for next-generation wireless communications, capable of mitigating severe propagation attenuation, enhancing spectral efficiency, and expanding signal coverage. This paper focuses on online millimeter-wave (mmWave) beam tracking for multi-RIS-assisted hybrid beamforming systems. We develop two novel beam tracking algorithms based on multi-agent deep reinforcement learning (DRL): a multi-agent deep deterministic policy gradient (MADDPG)-based algorithm for continuous-domain beam angle tracking and a multi-agent deep Q-network (MADQN)-based algorithm for codebook-based discrete-domain beam angle tracking. Both algorithms are designed to maximize the sum rate by jointly optimizing analog beamforming for the base station (BS) and reflection coefficients for multiple RISs in dynamic environments, leveraging historical information and without requiring current user position or channel information. After determining analog beamforming and RIS reflection coefficients, digital beamforming for the BS is constructed by estimating the end-to-end effective channel, which significantly reduces the overhead of channel estimation. Experimental results demonstrate that the proposed algorithms effectively adapt the analog beamformer and RIS reflection coefficients to account for user mobility, significantly outperforming existing benchmark schemes.

Multi-radio access technology (multi-RAT) enabled mobile edge computing (MEC) has emerged as a promising paradigm for supporting heterogeneous applications. However, efficiently managing resources for both ultra-reliable low-latency communications (URLLC) and enhanced mobile broadband (eMBB) services in large-scale networks remains challenging. In this paper, we investigate a joint optimization problem involving user association and bandwidth allocation in multi-RAT-enabled MEC systems. We propose a novel cooperative evolutionary framework operated based on the interplay between inner and outer agents to efficiently optimize large-scale networks. Extensive simulation results demonstrate that the proposed approach significantly outperforms the conventional single-RAT MEC system and several representative evolutionary computation algorithms.

UAVs are increasingly essential in wireless communication applications, such as internet of things (IoT) and sensor networks, due to their agile mobility. However, planning three-dimensional (3D) UAV trajectories over a continuous temporal-spatial domain remains challenging due to the computational complexity of non-convex optimization. This paper addresses UAV-assisted IoT data collection, aiming to minimize total energy consumption while considering UAV capabilities, heterogeneous IoT data demands, and 3D terrain. We propose a matrix-based differential evolution with constraint handling (MDE-CH), a computationally efficient algorithm for solving constrained non-convex optimization problems. Numerical results show that MDE-CH efficiently generates continuous 3D UAV trajectories, significantly reducing energy consumption and outperforming the conventional fly-hover-fly model for 2D and 3D trajectory planning.

Crowdsourced data refers to the information contributed by a large number of individuals, which may originate from various sources, including social media, online surveys, crowdsourcing tasks, etc. It is utilized for analysis and research across diverse scenarios. However, due to various subjective and objective factors, including the participants and sensing devices, the quality of the data collected for crowdsourcing tasks could be inconsistent. Therefore, how to filter out reliable information from the inconsistent data is crucial and difficult. Additionally, since participants consider their time and monetary costs, the crowdsourced datasets obtained are typically based on partial event observations, indicating a pronounced sparsity in the data. Current truth discovery methods struggle to adapt to datasets with varying levels of sparsity and lack effectiveness in evaluating and predicting sparse datasets that contain multiple judgments. In this article, we propose an adaptive hypergraph-based expectation-maximization (EM) truth discovery method for crowdsourced datasets with multiple judgments, named MHGEM (short for multidimensional-hypergraph EM). MHGEM leverages hypergraph topological metrics to model sparse datasets, enhancing the assessment of participant reliability and the prediction of truth for observed events. Experiments in both simulated and real-world scenarios demonstrate that MHGEM achieves higher predictive accuracy.

Renal cell carcinoma (RCC) is the most common type of kidney cancer in adults, its early detection and clinical diagnosis are crucial for reducing mortality. Deep convolutional neural networks (DCNNs) have demonstrated outstanding performance in medical image classification. Therefore, we selected the pre-trained EfficientNet-B2 as the backbone architecture for its lightweight and efficient characteristics. However, due to the implementation of privacy protection policies, acquiring sensitive medical data such as those related to kidney cancer has become increasingly challenging. Federated learning (FL) safeguards patient privacy by sharing only model parameters instead of raw data. Nevertheless, medical image data often exhibits non-independent and identically distributed (non-IID) characteristics, data sparsity, and imbalanced distributions, which adversely affect model performance. To address these issues, we propose an optimized federated learning framework, FedMilNet, which evaluates the data quality of each client and then utilizes dynamic curriculum learning. This approach arranges the sequence of client participation from easy to difficult based on data quality, gradually expanding the candidate frontier. Additionally, we determine “commanders” and “auxiliary teachers” based on the quality score, dynamically adjusting their weights as training progresses. FedMilNet achieved an accuracy of 62.71% on the ccRCC dataset, outperforming mainstream federated baselines by approximately 10%. On the Kaggle Kidney Cancer and CT KIDNEY DATASET, it reached high accuracies of 99.83% and 95.78%. By quantifying and utilizing client data quality in federated environments, and integrating curriculum-based client selection with dynamic knowledge distillation, our approach significantly enhances model generalization and accuracy under small-sample and non-IID conditions, advancing scientific progress in privacy-preserving and practical intelligent diagnosis of kidney cancer.

We present Keigo, a concurrency- and workload-aware storage middleware that enhances the performance of log-structured merge key-value stores (LSM KVS) when they are deployed on a hierarchy of storage devices. The key observation behind Keigo is that there is no one-size-fits-all placement of data across the storage hierarchy that optimizes for all workloads. Hence, to leverage the benefits of combining different storage devices, Keigo places files across different devices based on their parallelism, I/O bandwidth, and capacity. We introduce three techniques - concurrency-aware data placement, persistent read-only caching , and context-based I/O differentiation. Keigo is portable across different LSMs, is adaptable to dynamic workloads, and does not require extensive profiling. Our system enables established production KVS such as RocksDB, LevelDB, and Speedb to benefit from heterogeneous storage setups. We evaluate Keigo using synthetic and realistic workloads, showing that it improves the throughput of production-grade LSMs up to 4X for write- and 18X for read-heavy workloads when compared to general-purpose storage systems and specialized LSM KVS.

Learning multi-scale feature representations is essential for dense prediction tasks in autonomous driving. Most existing works are based on U-shaped architectures, where high-resolution representations are progressively recovered by connecting different levels of the decoder with low-resolution representations from the encoder. We observed that rich details from low-level representation and high semantic information from high-level representations are not fully utilized in the cross-stage fusion process. Additionally, current architectures often struggle to extract efficient discriminative feature along object boundaries. To address this issue, we propose CSNet, a generic cross-stage subtraction network that extracts spatial and semantic multi-scale representations through guided contextual feature. This approach allows fine-grained features to refine deeper layers, capturing discriminative high-resolution features while filtering out redundant information. Specifically, we introduce a cross-stage subtraction module (CSM), which consists of three sub-modules: 1) a Short Path Unit, focusing on capturing complementary adjacent information; 2) Medium Path Unit for effective middle-stages features aggregation; and 3) Long Path Unit for redundant information masking and long-range context modeling. Additionally, we propose the Semantic Guided Context Reasoning (SGCR) module to reason and model contextual relations between different subtraction units. CSNet demonstrates consistent performance gains across various semantic segmentation datasets. Our model, CSNet-M, achieves 82.2% mIoU on the Camvid dataset, while CSNet-S and CSNet-M attain 79.6% and 80.5% mIoU accuracy, respectively, on the Cityscapes dataset. These results show that the proposed CSNet has the potential for enhancing real-time semantic segmentation in autonomous driving applications, offering improved accuracy and efficiency in diverse urban scenarios. The source code for this work will be published athttps://github.com/mohamedac29/CSNet.

Recent federated unlearning studies mainly focus on removing the target client's contributions from the global model permanently. However, the requirement for accommodating temporary user exits or additions in federated learning has been neglected. In this paper, we propose a novel recoverable federated unlearning scheme, named RFUL, which allows users to remove or add their local model to the global one at any time easily and quickly. It mainly consists of two main components,i.e.,knowledge unlearning and knowledge recovery. In knowledge unlearning, the target contributions can be eliminated by training with mislabeled target data, while preserving the non-target contributions through distillation using the original model. In knowledge recovery, the forgotten contributions can be restored by training the target data using classification loss, while the non-target contributions are maintained through feature distillation and parameter freezing on the classifier. Both knowledge unlearning and recovery processes only require the participation of target data, guaranteeing the algorithm's practicality in federated learning systems. Extensive experiments demonstrate the significant efficacy of RFUL. For knowledge unlearning, RFUL matches state-of-the-art methods using only target data, achieving a runtime speedup of 3.3 to 8.7 times compared to retraining across various datasets. For knowledge recovery, RFUL exceeds state-of-the-art incremental learning methods by 5.02% to 29.97% in accuracy and achieves a runtime speedup of 1.8 to 4.4 times compared to retraining on different datasets.

Multi-radio access technology (multi-RAT) enabled mobile edge computing (MEC) has emerged as a promising paradigm for supporting diverse applications with heterogeneous service requirements. However, efficiently managing resources to accommodate both ultra-reliable low-latency communications (URLLC) and enhanced mobile broadband (eMBB) services remains challenging, especially in large-scale networks. In this paper, we investigate a joint optimization problem involving user association, task offloading, power and bandwidth allocation, and scheduling policies within a multi-RAT-enabled MEC system to efficiently address the heterogeneous demands of URLLC and eMBB services. We first formulate a generalized optimization problem and mathematically derive optimal power and task offloading strategies to reduce the search space. We then propose improved scheduling algorithms that sequentially update scheduling decisions based on arrival times at the edge server. Furthermore, we develop a matrix-based cooperative evolutionary computation framework with inner and outer agents to efficiently handle the large-scale optimization problem. Extensive simulation results demonstrate that our proposed approach significantly outperforms conventional scheduling methods and representative evolutionary algorithms.

Medical images often occupy large storage space and contain patient privacy or sensitive information, which makes them difficult and unsafe to be transmitted through the network. This paper proposed an adaptive compression-encryption scheme based on improved compressive sensing (CS) and deoxyribonucleic acid (DNA) coding-compression, which helps to solve the above problem. In the proposed scheme, first the original medical image was compressed to a floating point CS matrix using the discrete wavelet transform and the partial Hadamard matrix. The CS matrix was then normalized and rounded for matrix quantification. The result of quantification was fed to DNA fixed encoding and DNA run length coding for encryption and secondary compression. Finally the compressed-encrypted image was obtained after DNA dynamic encoding-decoding and regroups operation. The proposed scheme was tested against 9 images and proved to be effective in reducing the quantization errors and enhancing the compression performance. For instance, when the benchmark compression ratio (CR) is 0.5, the CR can be reduced by 5%(CR = 0.4453) ∼ 23%(CR = 0.2636), with the corresponding peak signal-to-noise ratio values consistently surpassing the benchmark. Furthermore, the execution of DNA compression and DNA dynamic encoding-decoding provided double guarantee for the algorithm’s security.

The growth of online social networks (OSNs) has become increasingly significant. Potential cloned accounts on these platforms raise serious concerns due to the risks they pose to user privacy and security. Previous works in the detection of cloned accounts on OSNs do not yield satisfactory results and lack consideration of the impact of missing attributes on the detection process. We propose cloned account detection with imputation framework for online social networks (CADIF-OSN) to accurately find potential cloned accounts on OSNs. This framework enables the accurate identification of potential cloned accounts on OSNs by leveraging their public profile information, even in cases where some of the information may not be accessible. The framework comprises four key components: 1) Fuzzy string matching with Levenshtein Distance that quickly generates suspicious account pairs by matching all the accounts' usernames and screennames; 2) An embedded method Doc2Vec that transforms all existing profile information of accounts into estimable vectors; 3) A HyperImpute model that imputes the missing information; and 4) A deep-forest model that is trained to detect cloned accounts. We evaluated our framework using a Twitter dataset consisting of 3,826 pairs of cloned accounts and 70,000 normal accounts. The evaluation results demonstrate that our framework significantly surpasses existing approaches in terms of Precision and F1-score.

Abstract With advancements in medical imaging technology, colour medical images make lesion diagnosis more intuitive. However, when transmitting these high‐capacity images, doctors and researchers must not only address the challenges of storage and transmission efficiency but also guard against unauthorized access and data security risks. To address these issues, a revolutionary approach for colour medical image compression encryption based on compressive sensing and deoxyribonucleic acid (DNA) coding operation is introduced in this study. Randomness and sparse optimizations are performed on three floating‐point matrices obtained through discrete wavelet and sparse transforms of plain colour medical images by employing position scrambling and reduced‐stiffness operations. Subsequently, the floating‐point matrices are measured and quantized to generate three 8‐bit integer matrices. Further, pixel‐by‐pixel DNA encoding, DNA‐base scrambling, DNA XOR, and DNA decoding operations are performed to achieve DNA base‐position scrambling and value diffusion. Finally, the regrouped bit planes help yield the compressed encrypted images. A comprehensive analysis of the proposed algorithm's encryption and decryption effectiveness, compression performance, and security was conducted. The results show that, with a compression ratio of 0.5, average PSNR = 42.7153 dB and average MSSIM = 0.9779, key space is , average entropy = 7.9986 bits, average histogram variance = 509.53, and the correlation coefficients are close to 0. Moreover, the algorithm shows some immunity to common cryptographic attacks, such as differential, known‐plaintext, noise, and occlusion attacks. Thus, the proposed algorithm addresses the challenges posed by the sensitive nature of patient information and limited storage space.

Federated learning (FL) stands as a promising distributed machine learning approach today, allowing model training on local clients without data sharing. However, varied contributions among clients often impact the global model’s performance, while FL confronts challenges like communication overhead, data heterogeneity, and privacy concerns. In this paper, we introduce a novel federated learning server aggregation algorithm: the Federated Learning Algorithm with Optimized Weight Aggregation via Particle Swarm Optimization Algorithm (AdpFedPSO). This approach dynamically adjusts client model contribution weights based on their performance and stability, aiming to enhance global model accuracy, convergence speed, and make model aggregation smarter and more adaptable. Through experimental validation on real datasets, we find that AdpFedPSO enhances accuracy by about 15% on the 0.6-Dirichlet MNIST dataset, 7.3% on the FashionMNIST dataset, and 13.4% on the CIFAR-10 dataset compared to the traditional FL aggregation method, FedAvg. These results indicate that AdpFedPSO not only enhances the accuracy of the global model but also expedites the model’s convergence speed. Additionally, it demonstrates resilience across various levels of client numbers, participation rates, and client heterogeneity, providing valuable reference and guidance for the further development of FL technology. Moreover, the concept of employing the particle swarm optimization algorithm for FL model aggregation also offers new insights and directions for future research.

Federated unlearning (FUL) is an emerging distributed machine learning paradigm which enables the removal or unlearning of specific training data effects from trained Federated Learning (FL) models. While current studies mostly focus on client-side FUL to address the “right to be forgotten”, and ignore the server's right to remove local models from the global model, particularly when clients are trained with low-quality data. In this paper, we introduce the Server-Initiated Federated Unlearning (SIFU) algorithm, devised to eliminate low-quality data from the global model. SIFU consists of two main components: (i) Identifying low-quality data: we develop a category-based method for quantifying low-quality data for each client and filter out clients containing such data. Datasets are then divided accordingly. (ii) Unlearning low-quality data: we employ gradient ascent training to counteract the adverse effects of low-quality data on local models. To minimize any bias introduced, we concurrently perform several batches of boosting training with good-quality data. SIFU could identify and promptly eliminate the impact of low-quality data on the FL global model while still preserving the benefits of good-quality data. Finally, extensive evaluations are conducted to verify the performance of SIFU with four different kinds of datasets and models. Results show that, compared to retraining from scratch, SIFU accelerates the speed of unlearning by 15× for small datasets (i.e., MNIST and FMNIST) and 20× for large datasets (i.e., CIFAR-10 and CelebA) without any degradation in accuracy, which also outperforms the state of the arts.

Analog DNA strand displacement circuits can be used to build artificial neural network due to the continuity of dynamic behavior. In this study, DNA implementations of novel catalysis, novel degradation and adjustment reaction modules are designed and used to build an analog DNA strand displacement reaction network. A novel adaptive linear neuron (ADALINE) is constructed by the ordinary differential equations of an ideal formal chemical reaction network, which is built by reaction modules. When reaction network approaches equilibrium, the weights of the ADALINE are updated without learning algorithm. Simulation results indicate that, ADALINE based on the analog DNA strand displacement circuit has ability to implement the learning function of the ADALINE based on the ideal formal chemical reaction networks, and fit a class of linear function.

DNA strand replacement technology has the advantages of simple operation which makes it becomes a common method of DNA computing. A four bit binary number Complementer based on two-domain DNA strand displacement is proposed in this paper. It implements the function of converting binary code into complement code. Simulation experiment based on Visual DSD software is carried out. The simulation results show the correctness and feasibility of the logic model of the Complementer, and it makes useful exploration for further expanding the application of molecular logic circuit.

Following the completion of the human genome project, a large amount of high-throughput bio-data was generated. To analyze these data, massively parallel sequencing, namely next-generation sequencing, was rapidly developed. DNA barcodes are used to identify the ownership between sequences and samples when they are attached at the beginning or end of sequencing reads. Constructing DNA barcode sets provides the candidate DNA barcodes for this application. To increase the accuracy of DNA barcode sets, a particle swarm optimization (PSO) algorithm has been modified and used to construct the DNA barcode sets in this paper. Compared with the extant results, some lower bounds of DNA barcode sets are improved. The results show that the proposed algorithm is effective in constructing DNA barcode sets.

The protein structure folding is one of the most challenging problems in the field of bioinformatics. The main problem of protein structure prediction in the 3D toy model is to find the lowest energy conformation. Although many heuristic algorithms h