This paper proposes a reputation-based federated learning framework to mitigate potential security threats in EEG signal classification. The framework is useful in the growing field of brain-computer interface technology, where EEG signal analysis is crucial. However, creating efficient learning models for such analysis is challenging due to the distributed nature of EEG data and the associated privacy and security concerns. The proposed framework addresses these issues by using the Federated Learning paradigm for privacy preservation and a reputation-based mechanism to reduce the impact of data poisoning attacks and identify compromised participants. Experimental results show the framework’s effectiveness in mitigating security threats.
Publication date: 5 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.01896
