KiFramework - A Federated Learning Platform

conference paper

English

This paper introduces a developed framework for federated learning on the Android mobile platform, specifically designed for crisis management applications. The framework primarily addresses challenges related to client communication during federated learning. To overcome the limitations of centralized server architectures, we have developed a novel communication protocol that ensures reliable communication during the learning process while facilitating efficient collaboration among multiple centralized servers. This protocol enables effective knowledge sharing and model training while maintaining data privacy and security. By leveraging this protocol, our framework enhances the performance and resilience of critical infrastructure systems on the Android platform, enabling real-time operations. Through this research, we contribute to improving crisis management applications by providing a comprehensive solution that optimizes communication and supports seamless collaboration across multiple centralized servers.

Android; communication protocol; federated learning; framework; machine learning

MICHÁLEK, J.; OUJEZSKÝ, V.; ŠKORPIL, V.

26. 11. 2024

IEEE

Spain

978-80-214-6295-3

2024 16th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT 2024)

6