FaceDCAPTCHA is designed to prevent face detection by automated bots through incorporating visual occlusions (black bars hiding face parts), applying visual distortions, and blending embedded images into the background. Adversarial learning is used to optimize the distortions that are applied to the CAPTCHA.
This CAPTCHA incorporates images of human faces, cartoons, and sketches on a multi-colored background. To solve FaceDCAPTCHA, users must tap or click on all the human face photographs.
Results
In testing, humans averaged 97%-100% accuracy in solving FaceDCAPTCHA images depending upon the distortions applied. Software face detection algorithms were completely unsucessful in solving this CAPTCHA. [1]
Demos
Publications
Journal Articles
- G. Goswami, B. M. Powell, M. Vatsa, R. Singh, and A. Noore, “FaceDCAPTCHA: Face Detection based Color Image CAPTCHA,” Future Generation Computer Systems, vol. 31, pp. 59–68, Feb. 2014. PDF Publisher's Website