False - Face Challenge Response Authentication Automated Test Using Anthropomorphic Images

Captcha is the reverse Turing test done by the machine to verify the human online automatically. This proposed work presents a method to use anthropomorphic (human outer cover morphed with monkey images) images for designing captchas. A set of false – face (morphed) images displayed in the grid with another group of images where the generated human-like pictures need to be as close as possible to human faces but not entirely. Such anthropomorphic captcha secure high score than other captchas while differentiating the human and machine. In the second level, clicked images are replaced by obfuscated, deformed, distorted and noised text images to decrease the machine recognition rate remarkably, compared to previous work while human recognition increased likewise. In addition to that, deep neural network model (proposed False-face model) trained with animal, human, morphed and alphabet images in a powerful way to classify the images. These images are tested with existing face classification algorithms like DeepFace, FaceNet, and TBE-CNN. This algorithm could not produce significant success in the prediction of anthropomorphic images. It helps to improve web security significantly while using this proposed captcha design.