MADPromptS: Unlocking Zero-Shot Morphing Attack Detection with Multiple Prompt Aggregation
This addresses a critical security challenge in face recognition systems by enabling more effective detection of morphing attacks without additional training, though it is incremental as it builds on existing foundation models.
The paper tackled the problem of Face Morphing Attack Detection (MAD) by proposing a zero-shot approach using CLIP without fine-tuning, focusing on aggregating multiple textual prompts per class to improve detection performance, with results showing substantial gains in zero-shot detection.
Face Morphing Attack Detection (MAD) is a critical challenge in face recognition security, where attackers can fool systems by interpolating the identity information of two or more individuals into a single face image, resulting in samples that can be verified as belonging to multiple identities by face recognition systems. While multimodal foundation models (FMs) like CLIP offer strong zero-shot capabilities by jointly modeling images and text, most prior works on FMs for biometric recognition have relied on fine-tuning for specific downstream tasks, neglecting their potential for direct, generalizable deployment. This work explores a pure zero-shot approach to MAD by leveraging CLIP without any additional training or fine-tuning, focusing instead on the design and aggregation of multiple textual prompts per class. By aggregating the embeddings of diverse prompts, we better align the model's internal representations with the MAD task, capturing richer and more varied cues indicative of bona-fide or attack samples. Our results show that prompt aggregation substantially improves zero-shot detection performance, demonstrating the effectiveness of exploiting foundation models' built-in multimodal knowledge through efficient prompt engineering.