Simulafed: an enhanced federated simulated environment for privacy and security in health

Abstract

Federated learning enables collaborative data analysis without the need to share sensitive information among participants, addressing privacy concerns in domains such as healthcare and finance. However, current federated simulation environments face challenges like limited flexibility in experiment configuration and difficulties ensuring data privacy. We present SimulaFed, a federated simulation environment based on a custom architecture that offers a personalised and configurable approach to data analysis using the Docker platform. SimulaFed allows researchers to create experiments tailored to their specific needs, ensures communication privacy, and incorporates various security and governance techniques. We demonstrate the effectiveness of SimulaFed through a real-world medical case, implementing and comparing two privacy-preserving federated algorithms for association rule mining: Tassa’s and Chahar’s algorithms. Our experiments show that while Tassa’s algorithm performs better in environments with a moderate number of participants due to lower computational and communication overhead, Chahar’s algorithm, though offering robust security through homomorphic encryption, suffers from efficiency limitations owing to high encryption and decryption costs. These findings provide valuable insights into the performance and limitations of existing algorithms, highlighting the need for more efficient methods in federated settings.