The Journal of Supercomputing最新文献

Fog computing is a revolutionary technology that, by expanding the cloud computing paradigm to the network edge, brings a significant achievement in the resource-constrained IoT applications in intelligent environments. However, security matters still challenge the extensive deployment of fog computing infrastructure. Ciphertext policy attribute-based encryption prepares a solution for data sharing and security preservation issues in fog-enhanced intelligent environments. Nevertheless, the lack of an effective mechanism to moderate the execution time of CP-ABE schemes due to the diversity of attributes used in secret key and access structure, as well as ensuring data security, practically restricts the deployment of such schemes. In this regard, a collaborative semantic model, including an outsourced CP-ABE scheme with the attribute revocation ability, together with an impressive AES algorithm relying on an ensemble learning system, was proposed in this study. The ensemble learning model uses multiple classifiers, including the GMDH, SVM, and KNN, to specify attributes corresponding to CP-ABE. The Dragonfly algorithm with a semantic leveling method generates outstanding and practical feature subsets. The experimental results on five smart building datasets indicate that the recommended model performs more accurately than existing methods. Also, the encryption, decryption, and attribute revocation execution time are significantly modified with the average time of 1.95, 2.11, and 14.64 ms, respectively, compared to existing works and conducted the scheme’s security.

As data grow exponentially, the demand for advanced intelligent solutions has become increasingly urgent. Unfortunately, not all businesses have the expertise to utilize machine learning algorithms effectively. To bridge this gap, the present paper introduces a cost-effective, user-friendly, dependable, adaptable, and scalable solution for visualizing, analyzing, processing, and extracting valuable insights from data. The proposed solution is an optimized open-source unsupervised machine learning as a service (MLaaS) framework that caters to both experts and non-experts in machine learning. The framework aims to assist companies and organizations in solving problems related to clustering and anomaly detection, even without prior experience or internal infrastructure. With a focus on several clustering and anomaly detection techniques, the proposed framework automates data processing while allowing user intervention. The proposed framework includes default algorithms for clustering and outlier detection. In the clustering category, it features three algorithms: k-means, hierarchical clustering, and DBScan clustering. For outlier detection, it includes local outlier factor, K-nearest neighbors, and Gaussian mixture model. Furthermore, the proposed solution is expandable; it may include additional algorithms. It is versatile and capable of handling diverse datasets by generating separate rapid artificial intelligence models for each dataset and facilitating their comparison rapidly. The proposed framework provides a solution through a representational state transfer application programming interface, enabling seamless integration with various systems. Real-world testing of the proposed framework on customer segmentation and fraud detection data demonstrates that it is reliable, efficient, cost-effective, and time-saving. With the innovative MLaaS framework, companies may harness the full potential of business analysis.

The exponential growth of Internet of Things (IoT) devices underscores the need for robust security measures against cyber-attacks. Extensive research in the IoT security community has centered on effective traffic detection models, with a particular focus on anomaly intrusion detection systems (AIDS). This paper specifically addresses the preprocessing stage for IoT datasets and feature selection approaches to reduce the complexity of the data. The goal is to develop an efficient AIDS that strikes a balance between high accuracy and low detection time. To achieve this goal, we propose a hybrid feature selection approach that combines filter and wrapper methods. This approach is integrated into a two-level anomaly intrusion detection system. At level 1, our approach classifies network packets into normal or attack, with level 2 further classifying the attack to determine its specific category. One critical aspect we consider is the imbalance in these datasets, which is addressed using the Synthetic Minority Over-sampling Technique (SMOTE). To evaluate how the selected features affect the performance of the machine learning model across different algorithms, namely Decision Tree, Random Forest, Gaussian Naive Bayes, and k-Nearest Neighbor, we employ benchmark datasets: BoT-IoT, TON-IoT, and CIC-DDoS2019. Evaluation metrics encompass detection accuracy, precision, recall, and F1-score. Results indicate that the decision tree achieves high detection accuracy, ranging between 99.82 and 100%, with short detection times ranging between 0.02 and 0.15 s, outperforming existing AIDS architectures for IoT networks and establishing its superiority in achieving both accuracy and efficient detection times.

Graph Neural Networks (GNNs) models, a current machine learning hotspot, have increasingly started to be applied in fraud detection in conjunction with user reviews in recent years. The accessible material is complicated and varied, the aggregated user evaluations cover a diverse range of topics, and erroneous information among vast amounts of user-generated content is typically rare. The review system is modeled as a heterogeneous network to address the issue of feature heterogeneity and uneven data distribution, and a new social theory-based graphical neural network model (SGNN) is suggested. The rich user behavior information in the heterogeneous network may be fully leveraged to acquire richer semantic representations for comments by integrating the hierarchical attention structure. Under the ensemble learning bagging framework, various distinct SGNN sub-models are combined. The sampling technique realizes the diversity aggregation of the base learners, which reduces the loss of useful information and improves the ability to identify bogus comments. According to testing results on real datasets from Amazon and YelpChi, the SGNN approach provides strong anomaly detection performance. It is demonstrated that the SGNN process has good robustness against fraudulent entities in the use of skewed distribution of data categories when compared to the existing approach.

UAV path planning poses the challenge of determining the most efficient route from an initial location to a desired destination, while considering mission objectives and adhering to various flight restrictions. This is a challenging optimization problem with high dimensionality that demands efficient path planning methods. To tackle the intricate UAV path planning problem within complex 3D environments, we propose an improved dung beetle optimizer (IDBO) for UAV path planning. Firstly, we formulate a cost function that converts the UAV path planning problem into a multidimensional function optimization problem, considering both trajectory restrictions and safety restrictions of the UAV. This enables us to effectively search for the optimal path. Secondly, we introduce a chaotic strategy to initialize the population, ensuring a comprehensive exploration of the solution space and enhancing population diversity. Additionally, we incorporate exponentially decreasing inertia weights into the algorithm, which improves convergence speed and exploration capability. Furthermore, to tackle the issue of decreasing population diversity during the late stages of convergence, we employ an adaptive Cauchy mutation strategy to enhance population diversity. Through simulation results, we demonstrate that IDBO achieves faster convergence and generates better paths compared to existing approaches in the same environment. These results demonstrate the remarkable efficacy of the proposed improved algorithm in effectively tackling the UAV path planning problem.

Introducing a novel approach for assessing connectivity in dynamic optical networks, we propose the quantum-driven particle swarm-optimized self-adaptive support vector machine (QPSO-SASVM) model. By integrating quantum computing and machine learning, this advanced framework offers enhanced convergence and robustness. Tested against a network simulation with 187 nodes and 96 DWDM channels, QPSO-SASVM outperforms traditional benchmarks such as LSTM, Naive method, E-DLSTM, and GRU. Evaluation using metrics such as signal-to-noise ratio, ROC curve, RMSE, and R² consistently demonstrates superior predictive accuracy and adaptability. These results underscore QPSO-SASVM as a powerful tool for precise and reliable prediction in dynamic optical network environments.

With the explosive growth of electronic information technology, mobile devices generate massive amounts of data and requirements, which poses a significant challenge to mobile devices with limited computing and battery capacity. Task offloading can transfer computing-intensive tasks from resource-constrained mobile devices to resource-rich servers, thereby significantly reducing the consumption of task execution. How to optimize the task offloading strategy in complex environments with multi-layers and multi-devices to improve efficiency becomes a challenge for the task offloading problem. We optimize the vertical assignment of tasks in a multi-layer system using deep reinforcement learning algorithms, which encompass the cloud, edge, and device layers. To balance the load among multiple devices, we employ the KNN algorithm. Subsequently, we introduce a task state discrimination method based on fuzzy control theory to enhance the performance of computing nodes under high load conditions. By optimizing task offloading policies and execution orders, we successfully reduce the average task execution time and energy consumption of mobile devices. We implemented the proposed algorithm in the PureEdgeSim simulator and performed simulations using different device densities to verify the algorithm’s scalability. The simulation results show that the method we proposed outperforms the methods in previous work. Our method can significantly improve performance in high-device density scenarios.

Trajectory prediction is highly essential for accurate navigation. Existing deep learning-based approaches always encounter serious performance degradation when facing shifted data or unseen scenarios. For learning transferable representations across different scenarios, the promising pretraining technique is applied to trajectory prediction tasks. However, relevant studies employ point-level masking mechanisms, which cannot capture local motion information across multiple time steps. Additionally, for trajectory data that couples multiple motion states, extracting the temporal dependencies within each state sequence remains highly challenging. To tackle this issue, we propose a channel-independent pretrained network via tokenized patching for efficient vehicle trajectory prediction, and it is composed of tokenized patch masking, channel-independent extractor (CiE), and state decoupling-mixing (SDM). Specifically, first of all, based on the designed tokenized patching scheme, TPM is established to represent local information and long-term relations in masked sequences. Then, through a series of weight-shared dense layers, CiE is designed to capture the individual dependencies among state sequences in an unsupervised pretraining manner. Moreover, by decoupling the complicated trajectory into pseudo-state representations, SDM is proposed to independently reconstruct the state sequences and further carry out representation mixing operations, to realize available trajectory predictions. Finally, extensive experiments show that our framework is effective and achieves the state-of-the-art performance on the INTERACTION and Argoverse2 datasets.

Public transportation systems play a vital role in modern cities, enhancing the quality of life and fostering sustainable economic growth. Modeling and understanding the complexities of these transportation networks are crucial for effective urban planning and management. Traditional models often fall short in capturing the intricate interactions and interdependencies in multimodal public transportation systems. To address this challenge, recent research has embraced multilayer network models, offering a more sophisticated representation of these networks. However, there is a need to explore and develop robustness analysis techniques tailored to these general multilayer networks to fully assess their complexities in real-world scenarios. In this paper, we employ a general multilayer network model to comprehensively analyze a real-world multimodal transportation network in Seoul, South Korea. We leverage a large volume of traffic data to model, visualize, and evaluate the city’s mobility patterns. Additionally, we introduce two novel methodologies for robustness analysis, one based on random walk coverage and the other on eigenvalue, specifically designed for general multilayer networks. Extensive experiments using the large volume of real-world data sets demonstrate the effectiveness of the proposed approaches.

Despite the enormous technical and financial advantages of cloud computing, security and privacy have always been the primary concerns for adopting cloud computing facilities, especially for government agencies and commercial sectors with high-security requirements. Homomorphic encryption (HE) has recently emerged as an effective tool in ensuring privacy and security for sensitive applications by allowing computing on encrypted data. One major obstacle to employing HE-based computation, however, is its excessive computational cost, which can be orders of magnitude higher than its counterpart based on the plaintext. In this paper, we study the problem of how to reduce the HE-based computational cost for general matrix multiplication, i.e., a fundamental building block for numerous practical applications, by taking advantage of the single instruction multiple data operations supported by HE schemes. Specifically, we develop a novel element-wise algorithm for general matrix multiplication, based on which we propose two HE-based general matrix multiplication algorithms to reduce the HE computation cost. Our experimental results show that our algorithms significantly outperform the state-of-the-art approaches of HE-based matrix multiplication.