Addressing biosignal data sharing security issues with robust watermarking

Abstract

One of the most important infrastructure require- ments in the domain of remote health monitoring BASNs is the secure collection and dissemination of the user's medical data. Data security desiderata in this application domain are not limited to ensuring the confidentiality and integrity of medical data that has been logged to a data sink. Requirements also arise from the need to provide the data owner (BAN user / patient) and the data consumers (healthcare providers, insurance companies, medical research facilities) secure control over the data as it is shared between these various stakeholders. Here, we study a robust watermarking technique to embed security information into biosignal data such that the semantic fidelity of the data is unaffected, while simultaneously ensuring that the watermark is not easily erased or corrupted by malicious data consumers. In doing so, we address three use-cases: proof of ownership, wherein the data owner can prove that she/he is the originator of the data; data tracking, wherein the data owner can trace unauthorized sharing of her/his biosignal data; and content authentication, wherein the data owner can prove whether the biosignal data has been maliciously altered. Based on experimentally collected datasets from a gait-stability monitoring BASN, we show that the embedding of 800 bit watermarks can be achieved robustly and effectively, with near-imperceptible changes to the signal waveform and no loss in the the signal's diagnostic quality.