Privacy Preserving Classification of EEG Data Using Machine Learning and Homomorphic Encryption

Data privacy is a major concern when accessing and processing sensitive medical data. A promising approach among privacy-preserving techniques homomorphic encryption (HE), which allows for computations to be performed on encrypted Currently, HE still faces practical limitations related high computational complexity, noise accumulation, sole applicability the at bit or small integer values level. We propose herein an encoding method that enables typical schemes operate real-valued numbers of arbitrary precision size. The evaluated two real-world scenarios relying EEG signals: seizure detection prediction predisposition alcoholism. supervised machine learning-based formulated, training using direct (non-iterative) fitting requires fixed deterministic number steps. Experiments synthetic data varying size complexity are determine impact runtime error accumulation. time models increases but remains manageable, while inference in order milliseconds. performance operating encoded comparable standard plaintext