Date of Publication :5th June 2025

Abstract: The threat of cyber-attacks, especially malware, is rapidly evolving and requires complex solutions that protect individual information from unauthorized access while providing high protection against malicious software. Federated Learning (FL) is a novel form of machine learning that enables model updates to be transferred among various clients without considering original data to be sent to a central hub. Several studies have investigated FL in cybersecurity; however, previous models present challenges associated with poisoning attacks, data heterogeneity, and no integration of sandbox for malware analysis. Based on this review thus critically discussed the current limitations on FL research in cybersecurity and possible solutions. Finally, they discuss the idea of combining Docker-based sandboxes with FL to solve these challenges, and they advocate for a feature-robust, privacy-preserving malware detection framework.

Reference :