An EEG Induced Model for Active User Authentication System

EEG-based Authentication Sytem Design

Project Overview

The presented authentication model continuously verifies a user’s identity throughout the user session while s/he watches a video or performs free-text typing on his/her desktop/laptop keyboard. The authentication model utilizes unobtrusively recorded electroencephalogram (EEG) signals and learns the user’s unique biometric signature based on his/her brain activity.

Datasets

The performance analysis of the proposed authentication system was performed on a publicly available dataset, BS-HMS Dataset. The dataset consists of brain signal data of 27 volunteer participants captured through Emotiv+ headset device while the users watch videos or type the free-text about the watched videos. Although the experimental analysis in this paper shows the efficacy of our authentication system while the users watch videos or type on their laptop/desktop keyboard, our results show viability for an EEG-based continuous authentication system for a wide range of activities.

Experimental Design

For our study, three different authentication models for each user were created: (i) Naive Authentication Model (ii) Subject-specific Features Model, and (iii) Global Optimal Features Model. All three models for each user were trained using Session I data. For the registration phase, training data sets were created using samples of the genuine users along with samples from randomly selected impostor users from the rest of the users in our dataset. For the verification phase, we adopted the same approach by creating testing sets for each user using their session II data. It was ensured that both the genuine and the imposter samples were balanced in terms of the number of samples. More details of this work is available @ Concealable Biometric-based User Authentication.

Related Works

Thinking Unveiled: An Inference and Correlation Model to Attack EEG Biometrics

An EEG-based authentication system that utlizes the time and frequency domain features extracted from the EEG data captured from 27 human subjects. The authors trained Linear Discriminant Analysis (LDA), Random Forest (RF), Neural Network (NN) and Support Vector Machine (SVM) algorithms for each user and their models achieved mean EER between 5.8% and 8.6%.

Inexpensive Brainwave Authentication: New Techniques and Insights on User Acceptance

This work presented an authentication system based on brain biometrics. The authentication was performed while the user performs five different cognitive and semantic tasks. The experimental analysis performed on 56 volunteer participants show that their authentication system could achieve an EER of 14.5%.

Neurokey: Towards a new paradigm of cancelable biometrics-based key generation using electroencephalograms

An EEG-based authentication system that takes input as DWT and DFT features to achieve an average half total error rate of 3.05%. The experiments were conducted on two datasets - Dataset I where the data was captured from 7 volunteer participants while they were performing different mental tasks and Dataset II from 120 subjects when the subjects were performing visually invoked tasks.