What's Old is New Again: Classical Dimensionality Reduction for Efficient Saliency-Guided Biometric Attack Detection
Researchers have developed a new method for acquiring saliency maps in biometric presentation attack detection (PAD) systems. This approach utilizes classical dimensionality reduction techniques, specifically PCA and LDA, to generate saliency maps directly from raw training data, eliminating the need for human annotation or domain-specific knowledge. The method has been tested across various PAD domains, including iris, synthetic face, and fingerprint, demonstrating its effectiveness and scalability by outperforming baseline and even state-of-the-art saliency methods without additional resource investment. AI
IMPACT This research offers a more efficient and scalable way to improve the robustness of biometric security systems by leveraging classical techniques for saliency map generation.