Concurrency and Computation-Practice & Experience最新文献

The growing demand for data sharing in domains such as health care highlights limitations in existing solutions, including low search efficiency, coarse-grained access control, and heavy decryption overhead on users. To address these challenges, this paper proposes a hierarchical searchable attribute-based encryption scheme with outsourced decryption for blockchain-based data sharing (OD-H-SABE). OD-H-SABE introduces a hierarchical attribute structure alongside an outsourced decryption mechanism that offloads computationally intensive bilinear operations to the cloud server. Consequently, users only need to perform a single lightweight operation to complete decryption, significantly alleviating the computational burden. Furthermore, the scheme integrates searchable encryption with multi-keyword aggregate hashing, enabling efficient search with constant complexity regardless of the number of keywords. Leveraging the transparency and immutability of blockchain, smart contracts verify the integrity of results returned from the cloud server, ensuring data security and trustworthiness throughout the sharing process. Theoretical and experimental analyses demonstrate that OD-H-SABE achieves notable advantages over traditional schemes in terms of security, search efficiency, and computational overhead. For example, compared to MKS-VABE and BEM-ABSE, OD-H-SABE reduces encryption and user-side decryption overhead by approximately 20% and 42%, respectively. This makes it a practical and lightweight solution for constructing secure and efficient blockchain-based data-sharing platforms.

Financial markets exhibit high levels of non-stationarity and structural heterogeneity, which pose significant challenges to reinforcement learning (RL)-based portfolio optimization methods. To address these challenges, this paper proposes a Structure-Aware Deep Reinforcement Learning (SADRL) framework for cross-market portfolio optimization. The proposed framework explicitly models market structural dynamics through a structure encoder that identifies latent market regimes, while a policy learner adapts investment strategies accordingly. This dual-level learning mechanism enables the model to generalize across heterogeneous markets and remain stable under regime shifts. Extensive experiments on multiple cross-market datasets demonstrate that SADRL achieves superior risk-adjusted returns and improved robustness compared with conventional RL-based baselines. These findings highlight the potential of structure-aware learning for developing intelligent and adaptive decision-making systems in financial markets.

Fog presents a substantial latent hazard to highway traffic safety, significantly impairing drivers' visibility, thereby elevating the risk of high-speed collisions. While driving, occlusion of multiple targets against complex backgrounds diminishes the detection rate of vehicle detectors. Many existing vehicle detection methods depend on bounding box representations for vehicle identification, limiting their capacity to provide accurate localization, particularly in foggy highway conditions. To enable early warnings of preceding vehicles in fog, this article proposes AG-YOLOv10n, a novel vehicle detection method for foggy environments. This approach improves the model's adaptability to fog-induced target features by replacing standard convolutional layers with AKConv and incorporating the GCAM gated convolutional attention module to enhance the extraction of locally salient information, thereby improving vehicle recognition accuracy in fog. Simultaneously, the DeepSORT tracking algorithm is enhanced, with AG-YOLOv10n replacing the traditional Faster R-CNN detector, and combined with the Kalman filter and Hungarian matching mechanism to achieve stable tracking of vehicle targets. The proposed method enhances the accuracy, recall rate, and average precision of the baseline model by 1.4%, 0.6%, and 1.1%, respectively, on the foggy vehicle dataset. The results demonstrate that the proposed method effectively improves detection accuracy, real-time performance, and system robustness while maintaining the model's lightweight nature, which holds significant practical application for highway fog driving safety.

In the context of security protection against false data injection attacks (FDIAs) in power grids, traditional federated learning effectively utilizes decentralized data resources for distributed training and achieves global collaboration. However, during the model aggregation process, it often overlooks or drowns out local sparse key features, significantly increasing the risk of missed detection of specific attack patterns. To address this issue, this paper proposes a personalized detection framework based on federated learning. Initially, the bidirectional transformer detection (BTD) model detection algorithm is deployed on the client side and trained on local data. Subsequently, through personalized federated learning, the client dynamically combines the weights of the global and local models to generate a personalized detection model. The framework employs a collaborative optimization mechanism of “global knowledge sharing and local feature adaptation” to effectively mitigate the feature drowning problem while strictly safeguarding data privacy. Compared to existing methods, this approach significantly enhances detection accuracy and robustness against differentiated attack patterns, thereby establishing a more reliable security defense system for smart grids.

Telecommunication fraud has escalated in complexity due to evolving adversarial strategies that exploit dynamic communication patterns, multimodal signals, and semantic manipulation. Recent developments in deep learning and graph-based modeling have shown promise; however, existing systems struggle to simultaneously capture temporal dependencies, relational feature structures, and linguistic nuances embedded in modern fraud activities. Addressing these limitations, this study proposes an Improved Graph Convolutional Network–Bidirectional LSTM (IGCN–Bi-LSTM) framework integrated within a unified signal-to-text and multi-perspective feature-engineering pipeline for high-accuracy fraud detection. The system begins by converting raw telecommunication signals into structured textual representations through a CNN-driven signal-to-text encoder, enabling the extraction of temporal–spectral patterns. These sequences are subsequently enriched through a comprehensive feature-engineering module that synthesizes linguistic markers, statistical descriptors, lexical indicators, and semantic embeddings. The hybrid IGCN–Bi-LSTM model then jointly learns higher-order relational dependencies among features and bidirectional temporal patterns, while an adaptive score-level fusion mechanism optimally weights model outputs for robust classification. Experiments were conducted using a high-quality synthetic Fraud Detection Transactions Dataset comprising 50,000 transactions with 21 heterogeneous attributes covering behavioral, contextual, financial, and security-related characteristics. Extensive preprocessing, normalization, and stratified data partitioning ensured reliable training of the hybrid model in a GPU-accelerated environment. The proposed model demonstrated substantial improvements over baseline methods by effectively capturing weakly correlated, high-dimensional features and rare-event patterns. Performance evaluation using precision-recall metrics confirmed the superiority of the IGCN–Bi-LSTM fusion, particularly in highly imbalanced scenarios where conventional accuracy metrics fail to reflect true detection capability.

The development of virtual platform frameworks has made it possible to perform early soft error analysis of more realistic multicore systems, that is, real software stacks and state-of-the-art ISAs. Because of the underlying frameworks' strong observability and simulation performance, more error/failure related data may be generated and collected in a reasonable amount of time even with complicated software stack setups. Parameters (i.e., features) that do not directly connect to the system soft error analysis must be filtered away when working with sizable failure-related data sets that come from several fault campaigns. In this regard, the paper proposes an assessment of multicore processor soft error reliability using BBRO-DNN and SSF-FIS models. At first, source code is converted into the executable code using LLVM compiler and applied over the Gem 5 virtual platform. Then, faults are injected into the fault injection module of the virtual platform. Profiling module analysis the faults and the reaction of the system and submits the report. The fault report is given into the proposed BBRO-DNN model for classifying the fault type. Finally, the system's reliability is evaluated using classified fault type. Experimental results are done by comparing the proposed and existing models to show the superiority of the developed model.

Cross-chain interoperability is a focal point in blockchain research. However, existing efforts predominantly concentrate on functional realization, such as asset transfer and message relaying, while often overlooking the critical dimension of performance. The effective implementation of these functions faces significant performance challenges, including data verification overhead, traceability delays, and inflexible contract execution. Precisely addressing this gap, this paper proposes a performance-optimized, relay-based architecture. Three dedicated core modules are introduced to address these performance bottlenecks: A Shared Data Life-cycle Management Module for efficient data governance, a Real-time Cross-chain Traceability Module for low-latency tracking, and a Dynamic Smart Contract Management Module for agile cross-chain logic execution. Implemented on BitXHub, our system demonstrates superior performance, successfully processing 937 out of 1000 transactions and achieving a latency of 6.7 ms under 800 concurrent requests. The framework's practical effectiveness is further validated through deployments in a cross-border seafood supply chain and a multi-party Deoxyribonucleic Acid (DNA) data sharing network, proving its value as a high-performance solution for complex real-world applications.

Electricity load data is usually significantly trending, cyclical, and stochastic in nature; it is also influenced by external factors such as weather, holidays, and socioeconomic activities. In addition, electricity loads usually have long-time dependencies, such as daily, weekly, and yearly periodicity. To address these challenges, we propose an electricity load forecasting method that combines PCA-PSO-Kmeans++ clustering with improved depthwise separable convolution. This article consists of two parts: data processing and electricity load forecasting. The data processing part of this includes seasonal decomposition of raw data, the Pearson correlation coefficient to select suitable exogenous variables, manual feature processing of raw electricity load data, and cluster analysis of feature-processed data. The electricity load forecasting part of the model is trained using an improved depthwise separable convolution that incorporates an attention mechanism and residual connection. Electricity load forecasting on datasets from the US and Nordic region. Experimental results show that clustering combined with improved depthwise separable convolution is more accurate and reliable in electricity load forecasting. Based on experimental results, we quantify the performance gain contributed by clustering relative to the strongest non-clustered baseline and demonstrate that clustering combined with the improved DSC further enhances accuracy.

Spider silk has exceptional mechanical properties, notably its high toughness, which is a measure of a material's ability to absorb energy before failure. Predicting toughness using physical, biochemical, and genetic features is a challenging task due to the nonlinear and multivariate interactions involved. This study presents a comprehensive machine learning framework to predict the toughness and the tensile strength of spider silk fibers using interpretable and high-performing models. A curated dataset with varied physicochemical and structural features was used to train, tune, and evaluate multiple machine learning models, including Decision Tree, support vector machines, Random Forest, Gradient Boosting, and XGBoost. Feature engineering steps introduced domain-specific constructs such as a toughness proxy and modulus transformations. Hyperparameter tuning was conducted via Bayesian optimization to enhance model performance. Among all tested models, the tuned XGBoost regressor achieved the highest predictive accuracy ($��{�R�}^{�2�}�=�0�.�855�$ and 0.765), outperforming all other models. Feature importance analysis highlighted several key predictors among Young's modulus, aligning with known biological mechanisms for both toughness and tensile strength. This work demonstrates how machine learning can be used not only for accurate prediction but also to uncover the underlying determinants of silk toughness. The proposed framework sets the stage for data-driven design of bioinspired synthetic fibers and represents a significant step toward the computational modeling of high-performance biomaterials.

Document level biomedical relation extraction aims to identify complex relationships between entity pairs in biomedical literature, which is crucial for the automation of medical knowledge applications. Existing methods face limitations when handling non-local and multi-layered semantic dependencies, making it difficult to effectively integrate global semantics with local interactions. The goal of this study is to propose a novel model to address this issue and enhance the ability to model complex dependencies. This paper proposes a new model that combines global dependency graphs with multi-level semantic information graphs (DMK). By utilizing a dual-graph collaborative mechanism, it integrates document-level contextual information to accurately model complex dependencies between entities. We introduce the KanChebConv convolutional layer based on the Kolmogorov–Arnold Network (KAN), replacing traditional linear weight matrices with learnable spline functions, thereby enhancing the model's ability to capture non-linear dependencies. We evaluated our model on the chemical–disease relation (CDR) dataset and the gene–disease relation (GDA) dataset. The results demonstrate that our model achieved the highest F1 score among the selected baselines on both datasets, thereby validating its robustness and competitiveness. Through the collaborative mechanism of global and local information and the innovative KAN convolutional layer, our model effectively improves the accuracy and robustness of document-level biomedical relation extraction, showcasing strong potential for practical applications.