Concerns about ensuring authorized access to sensitive data have grown as the digital era has progressed. Aside from standard identification methods, biometric qualities such as fingerprints, iris, facial characteristics, and, more recently, brain wave (based on electroencephalography – EEG) has piqued interest.
The current research on EEG-based authentication focuses on acute recordings in laboratory conditions utilizing high-end equipment with 64 channels and a high sampling rate. In a new study, researchers used a commercial dry-electrode EEG headset and chronic recordings on a group of 15 healthy persons to test the viability of EEG-based brain wave authentication in a real-world, out-of-laboratory setting.
A long short-term memory-based (LSTM) network is a recurrent neural network (RNN) dedicated to learning from time-sequence data. To decode the recordings in response to a multitask scheme consisting of performed and imagined motor tasks, the researchers used an LSTM network with bootstrap aggregating (bagging). They demonstrated that it outperformed the usual LSTM technique.