Faculty Advisor

Venkatasubramanian, Krishna Kumar

Abstract

The following report details the potential of using BVP and BCG waveforms as a biometrics means for user authentication using a head-mounted device. We trained a convolutional neural network to retrieve an optimal model that will allow for correct user authentication. We analyzed the results using ROC curves to determine the best performing models. The ROC curve analysis showed us that there were errors with our BVP results and therefore we moved on with only the BCG waveform. This research aimed to find alternative methods of authentication for users while considering the security of their passwords. The waveforms help us achieve this goal of authentication as users provide their subtle head movements which are much harder for adversaries to mimic than a typed password.

Publisher

Worcester Polytechnic Institute

Date Accepted

April 2019

Major

Computer Science

Major

Interdisciplinary

Project Type

Major Qualifying Project

Accessibility

Unrestricted

Advisor Department

Computer Science

Share

COinS