Federated Learning for Collaborative Cyber Defense

Abstract

With the increase in the complexity and number of cyber threats, security practitioners and defenders are looking for enhanced and robust security practices and solutions for effective cyber defense. Traditional cyber defense practices have been facing several challenges including data silos, data privacy concerns, and limited collaboration between organizations. In addition, the centralized machine learning models have their limitations, in terms of heterogeneity and formatting of datasets, for addressing distributed cyber threats. Federated learning has emerged as a decentralized machine learning approach with its potential for privacy preservation and contextualizing data for collaborative cyber defense. Federated learning has several characteristics, such as model aggregation, federated averaging, and differential privacy, which enable the collaborative training of machine learning models across distributed datasets.