Federated Learning Strategies for Privacy-Preserving Machine Learning Models in Cloud Computing Environments

Published: May 9, 2024

Language: Английский

---

Federated Learning-Based Predictive Traffic Management Using a Contained Privacy-Preserving Scheme for Autonomous Vehicles

Sensors, Journal Year: 2025, Volume and Issue: 25(4), P. 1116 - 1116

Published: Feb. 12, 2025

Intelligent Transport Systems (ITSs) are essential for secure and privacy-preserving communications in Autonomous Vehicles (AVs) enhance facilities like connectivity roadside assistance. Earlier research models used traffic management compromised user privacy exposed sensitive data to potential adversaries while handling real-time from numerous vehicles. This introduces a Federated Learning-based Predictive Traffic Management (FLPTM) system designed optimize service access within an ITS. Moreover, CPPS will provide strong security mitigate adversarial threats through state modelling authenticated permissions the integrity of vehicle communication networks man-in-the-middle attacks. The suggested FLPTM utilizes Contained Privacy-Preserving Scheme (CPPS) that decentralizes processing allows vehicles train local without sharing raw data. framework combines classifier-based learning technique with protect against invasions proposed model leverages Learning (FL) collaborative machine processes by allowing updates preserve privacy, enabling joint exposing It addresses key challenges such as high costs, impact attacks, time inefficiencies. Using FL, reduces costs 23.29%, mitigates effects 16.1%, improves 18.95%, achieving significant cost savings maintaining throughout process.

Language: Английский

---

Advanced transport systems: the future is sustainable and technology-enabled

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: April 24, 2024

Transport has always played a major role in shaping society. By enabling or restricting the movement of people and goods, presence absence transport services infrastructure historically been determining for cultures to connect, knowledge be shared, societies evolve prosper, or, contrast, decay fail. Since beginning twenty-first century, going through revolution worldwide. One primary goals sector is clear: it needs decarbonized become more sustainable. At same time, technological advances are toward smart societies. The Special Collection showcases some latest research towards sustainable technology-enabled transport.

Language: Английский

---

Adoption of K-means clustering algorithm in smart city security analysis and mythical experience analysis of urban image

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(3), P. e0319620 - e0319620

Published: March 10, 2025

Objective An information security evaluation model based on the K-Means Clustering (KMC) + Decision Tree (DT) algorithm is constructed, aiming to assess its value in evaluating smart city (SC) security. Additionally, impact of SCs individuals’ mythical experiences investigated. Methods analysis combination KMC and DT algorithms established. A total 38 are selected as research objects for practical analysis. The feasibility assessed using receiver operating characteristic (ROC) curve, performance compared with that Naive Bayes (NB), Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting (GBM) classification methods. Lastly, a questionnaire survey conducted obtain analyze SCs. Results (1) area under ROC curve significantly higher than 0.9 (0.921 vs. 0.9). (2) Compared NB LR algorithms, demonstrated true positive rate (TPR), accuracy, recall, F-Score, AUC-ROC, AUC-PR. metrics RF, SVM, GBM similar those KMC+DT model. (3) When attributes same, difference risk levels small, while when different, significant. (4) support rates various types new folk activities follows: offline shopping festivals (17.6%), New Year’s Eve celebrations (16.7%), Tibet tourism (15.6%), spiritual practices (16.2%), green leisure (16.0%), suburban/rural (15.8%). (5) High-risk cities (Grade A) showed stronger modern such leisure, low-risk (Grades C D) tended favor traditional cultural activities. Conclusion constructed this work capable effectively risks has value. good image mythological experience driving development cities.

Language: Английский

---

Enhancing Energy Efficiency of Sensors and Communication Devices in Opportunistic Networks Through Human Mobility Interaction Prediction

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1414 - 1414

Published: Feb. 26, 2025

The proliferation of smart devices such as sensors and communication has necessitated the development networks that can adopt device-to-device for delay-tolerant data transfer energy efficiency. Therefore, there is a need to develop opportunistic enhance efficiency through improved routing. A sensor device equipped with computing, communication, mobility capabilities opportunistically another device, either direct recipient or an intermediary forwarding third device. Routing algorithms designed aim increase probability successful message transmission by leveraging area information derived from historical forecast potential encounters. However, accurately determining precise locations mobile remains highly challenging necessitates robust prediction mechanism provide reliable insights into In this study, we propose incorporating random forest regressor (RFR) predict future location users, thereby enhancing routing RFR utilizes traces diverse users computing purposes. These predictions improve performance reduce bandwidth resource utilization during routine transmissions. To evaluate proposed approach, compared predictive against existing benchmark schemes, including Gaussian process, using real-world traces. University Southern California (USC) were employed underpin simulations. Our findings demonstrate significantly outperformed both process methods in predicting Furthermore, integration (D2D) traditional internet showed consumption reductions up one-third, highlighting practical benefits approach. contribution research it highlights limitations models develops new optimization energy-efficient overcome these limitations.

Language: Английский

---

Advanced Optimization Techniques for Federated Learning on Non-IID Data

Future Internet, Journal Year: 2024, Volume and Issue: 16(10), P. 370 - 370

Published: Oct. 13, 2024

Federated learning enables model training on multiple clients locally, without the need to transfer their data a central server, thus ensuring privacy. In this paper, we investigate impact of Non-Independent and Identically Distributed (non-IID) performance federated training, where find reduction in accuracy up 29% for neural networks trained environments with skewed non-IID data. Two optimization strategies are presented address issue. The first strategy focuses applying cyclical rate determine during while second develops sharing pre-training method augmented order improve efficiency algorithm case By combining these two methods, experiments show that CIFAR-10 dataset increased by about 36% achieving faster convergence reducing number required communication rounds 5.33 times. proposed techniques lead improved convergence, representing significant advance field facilitating its application real-world scenarios.

Language: Английский

---

Energy Consumption Monitoring and Prediction System for IT Equipment

Procedia Computer Science, Journal Year: 2024, Volume and Issue: 241, P. 272 - 279

Published: Jan. 1, 2024

This paper focuses on the monitoring and prediction of energy consumption IT equipment to make informed decisions in terms efficiency. The challenge with current systems lies their specialization, scalability integration complexities. To overcome these challenges, we propose a system for equipment. proposed solution combines an adaptable, cost-effiective energy-Efficient embedded device open source software service-oriented architecture (SOA), which offers flexibility capabilities, facilitating easy inclusion several workstation working from different environments. Several traditional Linear Regression (LR) models were evaluated using temporal window hour taking into account features. As result LR evaluation, it is established that Bayesian Ridge model was best since presented lowest error highest coefficient determination. Finally, two approaches predict consumption: Kernel Density Estimation (KDE)-based mechanism generate predictor variables order future model, KDE-based model. Numerical results show KDE measurements provides lower time response than based available dataset.

Language: Английский

---

Collaborative Federated Learning in Mobile Vehicle Clouds for Online Ride-Hailing Passenger Zones Recommendation

IEEE Internet of Things Journal, Journal Year: 2024, Volume and Issue: 11(22), P. 36646 - 36659

Published: June 27, 2024

Language: Английский

---

Resilient Privacy Preservation Through a Presumed Secrecy Mechanism for Mobility and Localization in Intelligent Transportation Systems

Sensors, Journal Year: 2024, Volume and Issue: 25(1), P. 115 - 115

Published: Dec. 27, 2024

An intelligent transportation system (ITS) offers commercial and personal movement through the smart city (SC) communication paradigms with hassle-free information sharing. ITS designs architectures have improved via technologies in recent years. The shared medium SCs is exposed to adversary risk, resulting privacy issues. Privacy issues impact contingent mobility localization of path. This paper introduces a novel resilient preserving (RPP) method presumed secrecy (PS) provide robust measure. progressive sessions preserved based on previous security depletion levels. interruptions traffic data-related are recurrently identified, re-handoffs recommended dodged transfer learning. empirical results indicate 25% reduction computational overhead 30% enhancement protection over conventional methods, demonstrating model's efficacy secure communication. Compared existing proposed approach decreases rates by 15% across varying densities, underscoring resilience high-interaction scenarios.

Language: Английский

---

Blockchain based intrusion detection in agent-driven flight operations

Multiagent and Grid Systems, Journal Year: 2024, Volume and Issue: 20(2), P. 161 - 183

Published: Aug. 12, 2024

Security and protection of the data is core objective every organization, but since cyber-attacks got more advanced than ever before, compromised often, resulting in financial loss, life or privacy breaches as its consequences. There must be a system that can deal with increasing number flight operations, which are numbers sophistication. Since we know traditional intrusion detection not capable enough to protect many human lives at stake an unfortunate corruption attack could give rise catastrophe. In this paper, proposed blockchain-based for operations framework data’s avoid operations. Blockchain only protects from also circumvents challenges faced by systems include trust consensus building between different nodes network enhance capability system.

Language: Английский

---