Reproduce “Robust Face Liveness Detection for Biometric Authentication using Single Image” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Poulami Raha, Yeongnam Chae

Researchers at Rakuten Group Inc. developed a lightweight Convolutional Neural Network (CNN) framework for robust face liveness detection against 2D presentation attacks. The system achieved an Average Classification Error Rate (ACER) of 2.7% on a custom dataset and performs real-time authentication within 1-2 seconds on a CPU.

Key insights

  • A carefully designed lightweight CNN can achieve high accuracy (2.7% ACER) against common 2D face presentation attacks while maintaining computational efficiency for CPU-based real-time deployment.
  • Leveraging contextual background information surrounding the face (e.g., paper or screen background vs. normal background) is an effective strategy for enhancing the robustness of spoof detection.
  • Developing a specialized, finely distributed dataset tailored to specific and prevalent 2D attack types is critical for training and validating anti-spoofing models for real-world application.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.