Pulse-to-Pair: A Heartbeat-based Authentication Method for Body Area Networks

Abstract

The expansion of Body Area Networks (BANs) and IoT-enabled healthcare has greatly improved medical monitoring and treatment. However, wireless BAN communications face significant security threats, with attackers potentially impersonating devices and disrupting networks. Current BAN security measures do not address the diverse user base, including older adults and cardiac patients. This thesis focuses on developing user-friendly and health-condition-sensitive security solutions for BANs. Cardiac signals can serve as the basis for a touch-to-access system that can be used for authentication and encryption purposes. They are a viable alternative to public key cryptography, eliminating the need for secure key storage and complex computations in resource-limited medical devices. Additionally, cardiac signals provide automated de-authentication when devices are no longer utilized. This work addresses the challenges in existing heartbeat-based authentication systems that fail to incorporate disease-induced abnormalities and different sensing mechanisms. We propose Pulse-to-pair, a cross-modality heartbeat-based authentication method that accommodates atypical signal properties, reduces false negatives in authentication, and provides lightweight security while enhancing usability. This thesis makes three contributions: designing a cross-modality biometric authentication and encryption system using Seismocardiography (SCG) and ECG signals, proposing an Inter-pulse Interval (IPI) alignment algorithm that substantially enhances the usability and efficiency, and investigating the impact of Valvular Heart Disease (VHD) on authentication. This is the first work to study VHDs from the perspective of biometric cryptography. A comprehensive experimental evaluation and security analysis are provided using ECG and SCG data from cardiac patients. We examine the impact of VHDs on the properties of heart signals and evaluate the suitability of various wavelets for processing these signals. Our analysis provides valuable insights into which wavelets perform best when dealing with signal changes induced by VHDs.