Mobile Brainwaves: On the Interchangeability of Simple Authentication Tasks with Low-Cost, Single-Electrode EEG Devices

Abstract

Biometric authentication, namely using biometric features for authentication is gaining popularity in recent years as further modalities, such as fingerprint, iris, face, voice, gait, and others are exploited. We explore the effectiveness of three simple Electroencephalography (EEG) related biometric authentication tasks, namely resting, thinking about a picture, and moving a single finger. We present details of the data processing steps we exploit for authentication, including extracting features from the frequency power spectrum and MFCC, and training a multilayer perceptron classifier for authentication. For evaluation purposes, we record an EEG dataset of 27 test subjects. We use three setups, baseline, task-agnostic, and task-specific, to investigate whether person-specific features can be detected across different tasks for authentication. We further evaluate, whether different tasks can be distinguished. Our results suggest that tasks are distinguishable, as well as that our authentication approach can work both exploiting features from a specific, fixed, task as well as using features across different tasks.

A Review of EEG-Based User Authentication: Trends and Future Research Directions

“…Feature extraction, Classification and Accuracy. With respect to the feature extraction task, the analyzed literature suggests the following: Waili et al [70] employed Wavelet transformation, Haukipuro et al [30], Di et al [20] and Nakamura et al [53] opted for PSD (Power Spectral Density), with the latter being combined also with Autoregressive (AR) models. La Rocca et al [60] also used AR models in their work.…”

“…Maiorana et al [49] selected Hidden Markov Models (HMM) and Curran et al [14] classified their data using Boosting techniques (XGBoost). There are also cases employing neural network based and/or deep learning architectures, with notable mentions being the work of Waili et al [70] and Haukipuro et al [30] who used a Multilayer Perceptron classifier (MLP), Kim et al [37] who used Functional Networks (FN) and finally, La Ma et al [48] and Shons et at al. [63] who relied on the usage of Convolutional Neural Networks (CNNs) for the classification task.…”

“…EEG complies with biometric trait standards, as each person produces unique and stable EEG signals [1,52], eligible for collection even via low-cost commercial devices and sensors [30]. Furthermore, they are suitable for mining, feature extraction, and classification [8].…”

A Biometric System Based on Single-channel EEG Recording in One-second

“…More recently, E. S. Haukipuro et al [7] explored the effectiveness of three different EEG authentication scenarios, which are resting, thinking about an image and moving one finger. They extracted features from Mel-Frequency Cepstral Coefficients (MFCC), and frequency power spectrum utilizing a multilayer perceptron classifier.…”

“…The main stages of any EEG-based biometrics system are scenarios of capturing the signal, Feature extraction and classification. Scenarios: according to the published studies, many different scenarios of capturing EEG signals have been used in the literature: resting with EC, resting with EO, responding to stimuli [2,3], performing a set of mental tasks (mental computations [4], imagined speech [5], and imagined movement [6]) where some mental tasks are more appropriate for EEG-based human recognition than others, and involve a particular movement such as moving a single finger [7]. The spatial distribution of the brain activations strongly dependent upon either the person's mental state or the activity performed during the acquisition [8].…”