Brain–Computer Interface Spellers for Communication: Why We Need to Address Their Security and Authenticity

摘要

Brain–Computer Interfaces (BCI) have witnessed significant research and development in the last 20 years where the main aim was to improve their accuracy and increase their information transfer rates (ITRs), while still making them portable and easy to use by a broad range of users. A recent review article on BCI-spellers by Rezeika and colleagues [ ] in the Brain Sciences journal provided an overview of different BCI paradigms, including P300, steady-state visual evoked potential (SSVEP), and motor imagery (MI), which are currently being used as communication channels for spelling. In these BCI-spellers, the output is provided as a graphical representation of letters, numbers, and symbols, or an audio output. The front-end of the BCI speller is the Graphical User Interface (GUI) which is very important according to Rezeika and colleagues [ ]. Indeed, the human-in-loop driving the BCI-speller receives feedback using the GUI (except in the case of motor imagery BCI), which therefore needs to be customized according to the residual function of the end-user. For example, SSVEP-based BCI will require a functional retina where the fidelity of the event-related potential (ERP) can be supplemented by auditory stimulation in a multimodal ERP-based BCI. However, the challenge is in the multi-modal sensor fusion, e.g., in optimizing the visual-to-auditory delay in multimodal spellers that combine visual and auditory stimulation [ ]. Such hybrid BCI systems may not only address subject-specific residual function but can also provide a more accurate and faster communication channel by combining different ERPs [ ]. An electroencephalogram (EEG)-based BCI speller has recently achieved ITRs of up to 5.32 bits per second and spelling rates up to 60 characters per minute [ ]. Also, soft electrode systems for EEG have been developed for continuous long-term use [ ]. With such advancement of the hardware and software for fast EEG-based communication channel for spelling, practical and naturalistic applications of BCI-spellers in daily living have become a reality, e.g., as a communication channel to send messages to the external world. In such practical applications, an important requirement for the BCI-speller will be to provide secure and authentic communication. In fact, user authentication (e.g., in the GUI) and resistance to attacks (e.g., in the BCI system) will have to be in-built for use as an official communication channel in real life.