Concurrency and Computation-Practice & Experience最新文献

Cloud computing is commonly utilized in remote contexts to handle user demands for resources and services. Each assignment has unique processing needs that are determined by the time it takes to complete. However, if load balancing is not properly managed, the effectiveness of resources may suffer dramatically. Consequently, cloud service providers have to emphasize rapid and precise load balancing as well as proper resource supply. This paper proposes a novel enhanced deep network-based load predictor and load balancing in cloud-fog services. In prior, the workload is predicted using a deep network called Multiple Layers Assisted in LSTM (MLA-LSTM) model that considers the capacity of virtual machine (VM) and task as input and predicts the target label as underload, overload and equally balanced. According to this prediction, the optimal load balancing is performed through a hybrid optimization named Osprey Assisted Pelican Optimization Algorithm (OAPOA) while taking into account several parameters such as makespan, execution cost, resource consumption, and server load. Additionally, a process known as load migration is carried out, in which machines with overload tasks are assigned to machines with underload tasks. This migration is applied optimally via the OAPOA strategy under the consideration of constraints including migration cost and migration efficiency.

In order to make networks more adaptable and flexible, software-defined networking (SDN) is an architecture that abstracts the many, easily distinct layers of a network. By enabling businesses and service providers to react swiftly to shifting business requirements, SDN aims to improve network control. SDN has become an important framework for Internet of Things (IoT) and 5G. Despite recent research endeavors focused on pinpointing constraints within SDN design components, various security attacks persist, including man-in-the-middle attacks, host hijacking, ARP poisoning, and saturation attacks. Overcoming these limitations poses a challenge, necessitating robust security techniques to detect and counteract such attacks in SDN environments. This study is dedicated to developing a method for detecting and mitigating control plane attacks within Software Defined Network Environments utilizing Deep Learning Algorithms. The study presents a deep-learning-based approach to identifying malicious hosts within SDN networks, thus thwarting unauthorized access to the controller. Experimental results demonstrate the effectiveness of the proposed model in host classification, exhibiting high accuracy and performance compared to alternative approaches.

Nowadays, intelligent transportation systems pay a lot of attention to autonomous vehicles it is believed that an autonomous vehicle improves mobility, comfort, safety, and energy efficiency. Making decisions is essential for the development of autonomous vehicles since these algorithms must be able to manage dynamic and complex urban crossings. In this research an optimal deep BiLSTM-GAN classifier to detect the movement of smart vehicles, initially the preprocessing stage is performed to decrease noise in the received data after that the essential regions are next be extracted in the region of interest (ROI) to make the right decision. The extracted data are forwarded to the GAN for road segmentation as well as the optimized deep BiLSTM classifier, which recognizes the traffic sign, simultaneously making it possible to do a modified Hough line-based maneuver prediction using the segmented information from the roads. Finally, the GAN determines the lane, and the BiLSTM predicts the traffic sign. The K-nearest neighbor (KNN)-based autonomous vehicle movement controllers are used to make the decision based on the predicted traffic sign and information about the lane. The proposed HSO algorithm was developed as the outcome of the common fusion of hawk and swarm optimization. Based on lane detecting achievements, at training percentage (TP) 90, the accuracy is 91.75%, Peak signal-to-noise ratio (PSNR) is 64.84%, mean square error (MSE) is 28.78, and mean absolute error (MAE) is 20.20, respectively, similarly based on the traffic sign prediction achievements at TP 90, the accuracy is 93.71%, sensitivity is 95.15%, specificity is 93.91%, and MSE is 28.78%, respectively.

Securing an e-commerce system using epidemic mathematical modeling with neural networks involves adapting epidemiological principles to combat the spread of misinformation. Just like how epidemiologists track the spread of diseases through populations, we can track the dissemination of fake news through online platforms. By modeling how fake news spreads, we gain insights into its propagation patterns, enabling us to develop more effective countermeasures. Neural networks, with their ability to learn from data, play a crucial role in this process by analyzing vast amounts of information to identify and mitigate the impact of fake news. One potential disadvantage of using epidemic mathematical modeling with neural networks to secure e-commerce systems is the complexity of the approach. The epidemic-based recurrent long short-term memory (E-RLSTM) technique addresses the complexity and evolving nature of fake news propagation by leveraging the strengths of recurrent neural networks (RNNs), specifically long short-term memory (LSTM) units, within an epidemic modeling framework. One advantage of using epidemic mathematical modeling with neural networks to secure e-commerce systems is its proactive nature. One significant finding in employing this approach is the ability to uncover hidden connections and correlations within the data. E-RLSTM stands out by capturing temporal dynamics and integrating epidemic parameters into its LSTM architecture, ensuring robustness and adaptability in detecting and combating fake news within e-commerce systems, outperforming other techniques in accuracy and performance. Description of the NSL-KDD dataset offers easy access to a valuable repository for benchmarking cyber security. Contained within are more than 120,000 authentic samples of cyber-attacks across 41 distinct categories, providing an excellent environment for testing intrusion detection systems.

Deep neural networks for medical image segmentation often face the problem of insufficient clean labeled data. Although non-expert annotations are more readily accessible, these low-quality annotations lead to significant performance degradation of existing neural network methods. In this paper, we focus on robust learning of medical image segmentation with noisy annotations and propose a novel noise-tolerant framework based on dual-strategy sample selection, which selects the informative samples to provide effective supervision information. First, we propose the first round of sample selection by designing a novel joint loss, which includes conventional supervised loss and regularization loss. To further select information-rich samples, we propose confidence-based pseudo-label sample selection from a novel perspective as the complement. The dual strategies are used in a collaborative manner and the network is optimized with mined informative samples. We conducted extensive experiments on datasets with both simulated noisy labels and real-world noisy labels. For instance, on a simulated dataset with 25% noise ratio, our method achieves segmentation Dice value with 90.56% $��\pm �$ 0.03%. Furthermore, increasing the noise ratio to 95%, our method still maintains a high Dice value of 73.85% $��\pm �$ 0.28% compared to other baselines. Extensive results have demonstrated that our method can weaken the effects of noisy labels on medical image segmentation.

An intrusion detection system (IDS) is valuable for detecting anomalies and unauthorized access to a system or network. Due to the black-box nature of these IDS models, network experts need more trust in systems to act on alerts and transparency to understand the model's inner logic. Moreover, biased models' decisions affect the model performance and increase the false positive rates, directly affecting the model's accuracy. So, maintaining Transparency and Fairness simultaneously in IDS models is essential for accurate decision-making. Existing methods face challenges of the tradeoff between fairness and accuracy, which also affects the reliability and robustness of the model. Motivated by these research gaps, we developed the Fair-XIDS model. This model clarifies its internal logic with visual explanations and promotes fairness across its entire lifecycle. The Fair-XIDS model successfully integrates complex transparency and fairness algorithms to address issues like Imbalanced datasets, algorithmic bias, and postprocessing bias with an average 85% reduction in false positive rate. To ensure reliability, the proposed model effectively mitigates the tradeoff between accuracy and fairness with an average of 90% accuracy and more than 85% fairness. The assessment results of the proposed model over diverse datasets and classifiers mark its model-agnostic nature. Overall, the model achieves more than 85% consistency among diverse classifiers.