Using EEG to Predict and Analyze Password Memorability

Brain-Computer Interfaces (BCIs) have given us insight into the human brain, and as sensors grow cheaper and smaller, and devices get more connected, more researchers from new domains are motivated to use BCIs. In this 19-participant lab study, we use off-the-shelf BCIs to investigate password memorability and recall. We record electroencephalogram (EEG) potentials collected by BCIs upon presenting passwords of different characteristics to participants while asking them to memorize these passwords, and then recall them. Features from the EEG signals are extracted in three domains: power spectrum from the frequency domain, statistics from the time domain, and wavelet coefficients from the time-frequency domain. Lasso feature selection method is used, and the selected parameters and feature subsets are submitted for classification with two classes, recalled and not recalled, based on the user's subsequent recall of the passwords. Results show discriminating features of EEG signals in the two different classes, achieving a classification accuracy of 88%. Our results indicate that it may be possible to predict subsequent password recall based on EEG activity during password presentation.