A Parametric Study of Arabic Text-Based CAPTCHA Difficulty for Humans

The Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) technique has been an interesting topic for several years. An Arabic CAPTCHA recently proposed serve Arab users. Since there have few scientific studies supporting a systematic design or tuning users, this paper aims analyze the text-based at parameter level by conducting experimental study. Based on results of study, we propose scheme with Fast Gradient Sign Method (FGSM) adversarial images. To evaluate security scheme, ran four filter attacks, which led success rate less than 5%. Thus, developed defensive method against adaptive attacks is standard evaluating defenses examples. This ensemble training, it gave accuracy result 41.51%. For usability aspect, conducted showed that can be solved humans in seconds 93.10%.