Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural Network-Powered Finger-Drawn Biometric Authentication

Maan Al Balkhi, Kordian Gontarska, Marko Harasic, Adrian Paschke

TL;DR

This work investigates finger-drawn digit authentication on touchscreens using neural networks, evaluating a lightweight shallow CNN, a modified Inception-V1 CNN, and two autoencoders for anomaly detection across data from 20 participants drawing 0–9 digits (2,000 samples each). The discriminative CNNs achieve about 89% authentication accuracy with AUC around 0.95, while autoencoders lag at approximately 75%, indicating strong potential for mobile, pattern-based biometrics with CNN-based methods. A key finding is that a shallow CNN provides nearly the same accuracy as a heavier model with far fewer parameters, enabling practical mobile deployment; autoencoders offer a complementary anomaly-detection approach with room for improvement. The results support finger-drawn symbols as a viable, secure biometric option that can complement existing pattern-based methods in MFA and continuous authentication on modern touchscreen devices, while guiding future work toward incorporating dynamic stroke features and broader symbol sets to boost impersonation resistance and usability.

Abstract

This paper investigates neural network-based biometric authentication using finger-drawn digits on touchscreen devices. We evaluated CNN and autoencoder architectures for user authentication through simple digit patterns (0-9) traced with finger input. Twenty participants contributed 2,000 finger-drawn digits each on personal touchscreen devices. We compared two CNN architectures: a modified Inception-V1 network and a lightweight shallow CNN for mobile environments. Additionally, we examined Convolutional and Fully Connected autoencoders for anomaly detection. Both CNN architectures achieved ~89% authentication accuracy, with the shallow CNN requiring fewer parameters. Autoencoder approaches achieved ~75% accuracy. The results demonstrate that finger-drawn symbol authentication provides a viable, secure, and user-friendly biometric solution for touchscreen devices. This approach can be integrated with existing pattern-based authentication methods to create multi-layered security systems for mobile applications.

Neural Network-Powered Finger-Drawn Biometric Authentication

TL;DR

This work investigates finger-drawn digit authentication on touchscreens using neural networks, evaluating a lightweight shallow CNN, a modified Inception-V1 CNN, and two autoencoders for anomaly detection across data from 20 participants drawing 0–9 digits (2,000 samples each). The discriminative CNNs achieve about 89% authentication accuracy with AUC around 0.95, while autoencoders lag at approximately 75%, indicating strong potential for mobile, pattern-based biometrics with CNN-based methods. A key finding is that a shallow CNN provides nearly the same accuracy as a heavier model with far fewer parameters, enabling practical mobile deployment; autoencoders offer a complementary anomaly-detection approach with room for improvement. The results support finger-drawn symbols as a viable, secure biometric option that can complement existing pattern-based methods in MFA and continuous authentication on modern touchscreen devices, while guiding future work toward incorporating dynamic stroke features and broader symbol sets to boost impersonation resistance and usability.

Abstract

This paper investigates neural network-based biometric authentication using finger-drawn digits on touchscreen devices. We evaluated CNN and autoencoder architectures for user authentication through simple digit patterns (0-9) traced with finger input. Twenty participants contributed 2,000 finger-drawn digits each on personal touchscreen devices. We compared two CNN architectures: a modified Inception-V1 network and a lightweight shallow CNN for mobile environments. Additionally, we examined Convolutional and Fully Connected autoencoders for anomaly detection. Both CNN architectures achieved ~89% authentication accuracy, with the shallow CNN requiring fewer parameters. Autoencoder approaches achieved ~75% accuracy. The results demonstrate that finger-drawn symbol authentication provides a viable, secure, and user-friendly biometric solution for touchscreen devices. This approach can be integrated with existing pattern-based authentication methods to create multi-layered security systems for mobile applications.

Paper Structure

This paper contains 26 sections, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Finger Recognition Methods
  4. Handwriting and Finger Writing Authentication Systems
  5. Neural Network Approaches for Authentication
  6. Neural Network Approaches for Authentication
  7. Strengths and Weaknesses of Existing Methods
  8. Methodology
  9. Data Set
  10. Experimental Design
  11. Training and Hyperparameter Optimization
  12. Neural Network Architectures
  13. Evaluation Methods
  14. Results
  15. Overall Performance Comparison
  16. ...and 11 more sections

Figures (9)

  • Figure 1: The architecture of the shallow CNN.
  • Figure 2: The architecture of the Convolutional Autoencoder.
  • Figure 3: The architecture of the Fully Connected Autoencoder.
  • Figure 4: Averaged ROC (left) and Discrimination Threshold Plot (right) over all participants for the shallow CNN.
  • Figure 5: Averaged ROC (left) and Discrimination Threshold Plot (right) over all participants for Mohapatra's CNN.
  • ...and 4 more figures