Brain-based Authentication: Towards A Scalable, Commercial Grade Solution Using Noninvasive Brain Signals.

Abstract

Here we report on a field test where we asked if it is feasible to deliver a scalable, commercial-grade solution for brain-based authentication given currently available head wearables. In this study, forty-nine (49) participants completed multiple sessions in their natural home environment over a single week. Participants used an off-the-shelf brain signal measuring headband to record their own brain activity while completing various tasks. Recording sessions were self-operated by the participants and unsupervised by any expert or technician to simulate real world use cases, while also contrasting common research approaches to this topic that rely on data from controlled laboratory conditions. Although brain signals have a non-stationary, complex nature, when participants watched rapidly presented images, our authentication system was able to successfully construct a unique and robust "brain ID" for each participant. Based on this brain ID, we developed a simplified brain-based authentication method that captures distinguishable information with reliable, commercial-grade performance from participants at their own homes. We conclude that noninvasively measured brain signals are ideal for use in biometric authentication systems, especially in environments where head wearables such as headphones or AR/VR devices are used as these devices offer a natural form factor for capturing participant brain ID continuously.