Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gaze Authentication: Factors Influencing Authentication Performance

Dillon Lohr, Michael J Proulx, Mehedi Hasan Raju, Oleg V Komogortsev

Abstract

This paper examines the key factors that influence the performance of state-of-the-art gaze-based authentication. Experiments were conducted on a large-scale, in-house dataset comprising 8,849 subjects collected with Meta Quest Pro equivalent hardware running a video oculography-driven gaze estimation pipeline at 72~Hz. State of the neural network architecture was employed to study the influence of the following factors on authentication performance: eye tracking signal quality, various aspects of eye tracking calibration, and simple filtering on estimated raw gaze. This report provides performance results and their analysis.

Gaze Authentication: Factors Influencing Authentication Performance

Abstract

This paper examines the key factors that influence the performance of state-of-the-art gaze-based authentication. Experiments were conducted on a large-scale, in-house dataset comprising 8,849 subjects collected with Meta Quest Pro equivalent hardware running a video oculography-driven gaze estimation pipeline at 72~Hz. State of the neural network architecture was employed to study the influence of the following factors on authentication performance: eye tracking signal quality, various aspects of eye tracking calibration, and simple filtering on estimated raw gaze. This report provides performance results and their analysis.

Paper Structure

This paper contains 36 sections, 3 equations, 2 figures, 10 tables.

Table of Contents

  1. INTRODUCTION
  2. PRIOR WORK
  3. METHODOLOGY
  4. Dataset
  5. Data Processing
  6. Network Architecture
  7. Model Training
  8. Model Evaluation
  9. Performance Metrics
  10. EXPERIMENTAL DESIGN
  11. Authentication Scenarios ( $\times$ 2)
  12. Calibration Parameters during Training ( $\times$ 2)
  13. Signal Quality ( $\times$ 2)
  14. Axis of operation ( $\times$ 3)
  15. Epochs and Minibatch Size ($\times$ 2)
  16. ...and 21 more sections

Figures (2)

  • Figure 1: Overview of the embedding model’s DenseNet-based network architecture lohr_ekyt. The input is 360 time steps (5 seconds @ 72 Hz), where each time step has 4 or 8 features (yaw/pitch velocity $\times$ left/right eye $\times$ optical/visual axis), depending on whether we use one or both axes. The output is a 128-dimensional embedding.
  • Figure 2: Diagram of which set of calibration parameters is used to estimate visual axis during enrollment and each verification scenario for both genuine and impostor verification attempts. Alice is enrolled in the system. At verification time, both Alice and an impostor Bob claim to be Alice.