Striking a balance between human usability and machine resistance has long been a challenge for CAPTCHA design. Many recent solutions rely on complex, resource-intensive techniques that often frustrate users. To address this, we exploit a lesser-known vulnerability in deep learning vision models, namely their difficulty in recognizing objects when texture conflicts with shape or context. Based on this insight, we introduce a novel CAPTCHA challenge design strategy using texture blending, which subtly disrupts machine perception while remaining clear to humans. Experiments and user studies show that our method offers strong robustness against automated attacks without sacrificing usability.
Camouflaged in Patterns: Designing Robust and User-Friendly CAPTCHA Challenges via Texture Blending
Moradi M.;Moradi M.;Palazzo S.;Spampinato C.
2025-01-01
Abstract
Striking a balance between human usability and machine resistance has long been a challenge for CAPTCHA design. Many recent solutions rely on complex, resource-intensive techniques that often frustrate users. To address this, we exploit a lesser-known vulnerability in deep learning vision models, namely their difficulty in recognizing objects when texture conflicts with shape or context. Based on this insight, we introduce a novel CAPTCHA challenge design strategy using texture blending, which subtly disrupts machine perception while remaining clear to humans. Experiments and user studies show that our method offers strong robustness against automated attacks without sacrificing usability.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
