Recent Advances in Privacy-Preserving Cybersecurity Using Homomorphic Encryption and Federated Learning

Abstract

Cybersecurity threats and the need to secure sensitive data have grown significantly as connected devices, IoT, cloud computing, and digital services have all exploded. This review examines the most recent developments in privacy-preserving cybersecurity, with a particular emphasis on Federated Learning (FL) and Homomorphic Encryption (HE). Digital signatures, hash functions, zero-knowledge proofs, Secure Multi-party Computation, and Denial-of-Service, Malware, Phishing, SQL Injection, Session Hijacking, Man-in-the-Middle, and Cross-Site Scripting are some of the methods covered. There are three main varieties of HE covered in the study: Partial Homomorphic Encryption, Somewhat Homomorphic Encryption, and Fully Homomorphic Encryption. HE is a method for computing without decrypting the data. Additionally, FL is covered as a decentralized ML strategy that uses data retention at the local level and updates models through distribution, similar to Horizontal, Vertical, and Federated Transfer Learning. Issues with data heterogeneity, straggler effects, privacy protection, and security are also covered, along with other FL difficulties. Applications in the healthcare, financial, internet of things (IoT), edge computing, and industrial sectors are highlighted, and HE and FL are shown to provide more privacy in new areas of application as well as safer, more collaborative, and more efficient cybersecurity solutions.