Original scientific paper

https://doi.org/10.17559/TV-20240528001690

Federated SignalGAN: Privacy-Preserving Collaborative Brain Signal Processing for Enhanced Diagnostic Accuracy

N. Deepa ; Kalasalingam Academy of Research and Education, Krishnankoil, Tamil Nadu, India *
R. Sumathi ; Kalasalingam Academy of Research and Education, Krishnankoil, Tamil Nadu, India

* Corresponding author.

Abstract

As the demand for enhanced privacy in collaborative brain signal processing intensifies, this research presents a robust federated learning framework. Collaborative signal analysis necessitates data pooling across institutions, emphasizing the critical need for privacy preservation. "Federated SignalGAN" is an innovative algorithm that unites Generative Adversarial Networks (GANs) with the core principles of federated learning. The adversarial training objective of GANs is to simultaneously train the generator and discriminator. This is formulated as a minimax game, with the generator striving to produce synthetic data that are indistinguishable from real ones, while the discriminator endeavors to become more adept at distinguishing between real and synthetic data. Federated Learning is a distributed machine learning paradigm to train a global prototype that minimizes a specific loss function while respecting the data privacy constraints of each participating institution. Federated brain SignalGAN stands out by introducing a unique approach for generating synthetic signal data, thereby eliminating the necessity for direct sharing of sensitive information. This research employs comprehensive simulation analysis to rigorously assess the performance of Federated brain SignalGAN. Key simulation metrics, including diagnostic accuracy, data privacy preservation, convergence rate, and GPU utilization are used to evaluate the effectiveness of this framework using open data repositories and real-time brain signal processing datasets. The implications of this research are profound, emphasizing the pivotal role of privacy-preserving federated learning frameworks in signal processing. By introducing a novel algorithm designed to meet the unique challenges of collaborative signal analysis, this research makes a substantial contribution to secure and accurate signal detection in distributed environments. The adoption of Federated brain SignalGAN is pivotal for ensuring data confidentiality while enabling effective multi-institutional signal analysis.

Keywords

federated learning; generative adversarial networks (GANs); privacy-preserving signal processing; collaborative brain signal analysis; synthetic data generation

Hrčak ID:

328647

URI

https://hrcak.srce.hr/328647

Publication date:

27.2.2025.

Visits: 746 *