Armoring Motor Imagery EEG Systems: A PSNR Optimized Differentially Private Topographic Mapping Mechanism

The rapid proliferation of Brain-Computer Interfaces (BCIs) has introduced critical security vulnerabilities in neural data processing pipelines. While Motor Imagery (MI)-based EEG systems enable direct neural control of assistive devices, current frameworks remain susceptible to spectral leakage attacks, adversarial perturbations, and signal spoofing threats. To address these challenges, we propose a privacy-preserving Electroencephalogram (EEG) processing framework that integrates three novel components: A differentially private topographic mapping mechanism achieving a reduction in μ-rhythm reidentification risk while maintaining signal fidelity loss; An encrypted epoch fusion protocol utilizing lattice-based homomorphic encryption to ensure 79.50% average classification accuracy under adaptive white-box attacks; A PSNR-optimized dual metric system that enhances inter-class separability while compressing intra-class variance. Experiments conducted on BCI Competition IV-2a datasets demonstrate that our method establishes new benchmarks for trustworthy neural interface development.