A Research Review on Challenges of Federated Machine Learning

• Published in: International Conference on Computing Communication Control and Automation (Online) pp. 1 - 7
• Main Authors: Adsure, Sharad, Devare, Manoj
• Format: Conference Proceeding
• Language: English
• Published: IEEE 23.08.2024
• Subjects: Computational modeling; Data models; Data privacy; DP Differential Privacy; etc; Federated learning; Privacy; Scalability; Security; Stress; Training; Transfer learning
• ISSN: 2771-1358
• DOI: 10.1109/ICCUBEA61740.2024.10775251

Federated Learning (FL) offers a decentralized approach to model training, preserving data privacy, scalability, and diversity compared to centralized methods. By distributing computational tasks across devices and facilitating localized training, FL minimizes communication burdens and ensures secure handling of sensitive data. This innovative technique enables organizations to train AI models collaboratively on distributed data without central storage or sharing, thereby preserving privacy and advancing machine learning capabilities. Preserving privacy in federated learning (FL) presents multi-faceted challenges, including diverse data distributions, communication security risks, and potential privacy leaks during aggregation. Maintaining model efficacy while implementing effective privacy-preserving techniques is crucial, necessitating a careful balance between privacy and performance. Additionally, ad- dressing vulnerabilities to model inversion attacks and ensuring participant compliance with privacy protocols adds complexity to FL systems. To overcome these challenges, innovative solutions and robust security measures are imperative to safeguard user privacy and uphold FL model efficacy while prioritizing fairness and privacy preservation.