Image Spoofing in Biometrics: Evolving Threats, Detection Strategies, and Future Directions

Abstract

Over the past few decades, biometric technology has advanced significantly, beginning with the earliest studies on voice and facial recognition and continuing to this day with a variety of highly accurate systems. These modalities range from widely deployed ones like fingerprint, face, or iris to less common modalities like handwriting or signatures. Image spoofing poses a significant threat to security systems that rely on visual data for authentication. This survey evaluates a number of the most widely used strategies in each field, looking at how they work, their advantages, and any potential drawbacks. We will wrap up by summarizing the current state of the art, highlighting the unresolved issues, and providing an overview of potential future paths for this study. Additionally, we will present an organized future research roadmap, and identify unresolved obstacles. This survey makes three distinctive contributions: (i) a unified cross-modal taxonomy classifying attacks by type, modality, and detection strategy; (ii) a critical analysis of cross-domain generalization gaps, explaining why certain detection approaches outperform others across datasets; and (iii) integrated coverage of emerging generative-AI threats including GAN-based and diffusion-model-generated spoofs alongside practical deployment considerations such as computational cost and edge-device suitability.