Cryptographic Key Generation using Electroencephalograms

Garima Bajwa, University of North Texas
Ram Dantu, University of North Texas

Background. Brain waves (Electroencephalograms: EEG) can provide unconscious continuous human authentication for the intended system.

Aim. The intent of this work is to reliably generate a unique and repeatable key from a user’s EEG signals which is resistant to cryptanalysis, eves dropping and coercion attack even against an adversary who obtains all the information regarding the system.

Method. Seven subjects complete five mental tasks for repeated trials with data recorded from six electrodes. Features for the first step Subject Authentication are obtained from each task using power of the frequency domain data. Second step constituting the Neurokey generation involves feature extraction from the global and local normal distribution fit for the same frequency data.

Results. An accuracy of hundred percent is observed for subject classification using different classifiers for the tasks. Unique key is established for each subject. However, repeatability is in question considering the false acceptance and rejection rates.

Conclusions. A consistent unique key for each subject can be obtained using EEG signals by collecting data from distinguishable mental activities. Researchers are encouraged to establish stronger mapping from EEG feature space to binary codewords for stable key bits generation.