Secure Stochastic ECG Signals Based on Gaussian Mixture Model for $e$-Healthcare Systems

The blood circulation system in a human body provides a unique and natural trust zone for secure data communications in wireless healthcare systems such as body area networks. Unfortunately, biometric signal authentication using physiological attributes in wireless healthcare has not been extensively studied. In this paper, we propose a data authentication approach utilizing electrocardiography (ECG) signal patterns for reducing key exchange overhead. The major contribution of this research is to apply stochastic pattern recognition techniques in wireless healthcare. In the proposed approach, the inter-pulse interval (IPI) signal pattern at transmitter side is summarized as a biometric authentication key using Gaussian mixture model (GMM). At the receiver side, a light-weight signature verification scheme is adopted that uses IPI signals gathered locally at the receiver. The proposed authentication scheme has the advantage of high sample misalignment tolerance. In our earlier work, we had demonstrated the concept of stochastic authentication for ECG signal, but the signature verification process and GMM authentication performance under time synchronization and various sample points were not discussed. Here, we present a new set of analytical and experimental results to demonstrate that the proposed stochastic authentication approach achieves a low half total error rate in ECG signals verification.