A Survey on Privacy-Preserving Communication Frameworks in Machine Learning for Cybersecurity Threat Detection

This survey covers the new trend of privacy-preserving communication frameworks in machine learning in the FL, DP and SMPC for cyber threat detection domains. The specific research performed between the year of 2020 and 2024 has been surveyed and included in this paper. The main contribution is monitoring cybersecurity threats using privacy-preserving frameworks in distributed systems such as IoT and industrial networks. 40 research papers have been reviewed and categorized to several related themes such as FL implementations, privacy-preserving methods, and cybersecurity threat detection. The relevance of privacy-preserving frameworks for the real-world cybersecurity scenario is illustrated with dataset CISA 2022 Vulnerability Dataset. The study of privacy-preserving communication shows that FL can be combined with DP and SMPC to achieve data privacy preservation while deduction accuracy of detecting cyber threats can be well-maintained. Moreover, the research gaps and challenges of scalability, communication overhead have been addressed and identified to facilitate the future research and development.