Camila, Torres and Diego, Rojas and Isabella, Martínez (2024) Federated Learning for Privacy-Preserving Cybersecurity Applications. American Journal of Technology Advancement, 1 (8). pp. 98-113. ISSN 2997-9382

Text Federated Learning for Privacy-Preserving Cybersecurity Applications.pdf Download (349kB)

Abstract

The increasing sophistication of cyberattacks and the exponential growth of sensitive digital data have intensified the demand for advanced, privacy-preserving cybersecurity solutions. Traditional centralized machine learning approaches require aggregating large volumes of data into a single repository, raising significant concerns about data privacy, regulatory compliance, and vulnerability to breaches. Federated Learning (FL) has emerged as a transformative paradigm that enables collaborative model training across decentralized devices and organizations without sharing raw data. This privacy-preserving architecture is particularly relevant in sectors such as finance, healthcare, and critical infrastructure, where regulatory frameworks like GDPR and HIPAA impose strict data handling requirements. Recent empirical studies demonstrate that FL-based intrusion detection systems can achieve detection accuracies exceeding 92% on benchmark datasets (e.g., CICIDS2017), while reducing data exposure risks compared to centralized approaches. Moreover, industry pilots highlight FL’s scalability, with Google successfully deploying it to over 1 billion mobile devices for security and personalization tasks. Despite its promise, FL faces challenges including communication overhead, model poisoning, and heterogeneity of local datasets. This paper investigates the potential of federated learning in cybersecurity applications, focusing on intrusion detection, malware classification, and IoT security. It further explores techniques such as differential privacy and secure multi-party computation to enhance resilience against adversarial manipulation. The findings underscore that federated learning not only advances threat detection capabilities but also aligns cybersecurity practices with the pressing need for data confidentiality, making it a viable strategy for privacy-preserving, collaborative defense in the evolving digital threat landscape.

Item Type:	Article
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Postgraduate > Master's of Islamic Education
Depositing User:	Journal Editor
Date Deposited:	27 Sep 2025 12:01
Last Modified:	27 Sep 2025 12:01
URI:	http://eprints.umsida.ac.id/id/eprint/16392

Actions (login required)

View Item