Privacy-Preserving Techniques in Big Data: Enhancing Security through Federated and Different Learning
BigDataStream Mining · 30.06.2025 · 1
DOI: 10.70023/sahd/b121 ↗Kurz gefasst
Fachartikel in der Zeitschrift BigDataStream Mining, online erschienen am 30.06.2025. Verfasst von Fatima Al Neyadi und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The rapid growth of big data analytics in areas like healthcare, finance, smart cities, and IoT systems has brought to light important problems with data privacy, security, and following the rules. When sensitive information is put into one place, as traditional centralized learning systems require, it is easier for data breaches, unauthorized access, and privacy inference attacks to happen. These challenges underscore the necessity of formulating robust privacy-preserving techniques to enable collaborative data analytics while safeguarding the integrity of sensitive information. This work presents a privacy-preserving large data solution that enhances security while minimally impacting analytical performance. It has privacy methods that use Different (Differential) Learning and Federated Learning (FL). The suggested methodology ensures the continuous preservation of the data owner's environment, as federated learning (FL) enables several remote clients to train a shared global model locally. It can improve privacy-enhancing differential learning methods by using safe aggregation techniques and adding calibrated noise to local model updates. This helps lower the chances of attacks like model inversion and membership inference. Tests on distributed large data settings have proven that the proposed technique provides strong privacy protections without hurting model stability or prediction accuracy. Compared to centralized learning, the approach makes the system much more resistant to attacks and significantly lessens the risk of data leakage, all with very little performance loss. Also, FL's decentralized nature makes it easier to follow data protection standards and makes it easier to grow. Finally, federated and differential learning may assist you get safe big data analytics that don't violate privacy. The suggested framework is good for apps that need to be secure and follow rules since it achieves a good balance between privacy and data usefulness. Adaptive privacy budgets and hybrid trust mechanisms may be areas for future study focused on enhancing system resilience.
Zitieren
APA
Neyadi, F. A., & Mansoori, A. A. (2025). Privacy-Preserving Techniques in Big Data: Enhancing Security through Federated and Different Learning. BigDataStream Mining, 1. https://doi.org/10.70023/sahd/b121
Vancouver
Neyadi FA, Mansoori AA. Privacy-Preserving Techniques in Big Data: Enhancing Security through Federated and Different Learning. BigDataStream Mining. 2025:1. doi:10.70023/sahd/b121
BibTeX
@article{Neyadi2025,
author = {Neyadi, Fatima Al and Mansoori, Ahmed Al},
title = {Privacy-Preserving Techniques in Big Data: Enhancing Security through Federated and Different Learning},
journal = {BigDataStream Mining},
year = {2025},
pages = {1},
doi = {10.70023/sahd/b121}
}
EUVIMED