Style matching CAPTCHA: match neural transferred styles to thwart intelligent attacks

Completely automated public turing test to tell computers and humans apart (CAPTCHA) is widely used prevent malicious attacks on various online services. Text- image-CAPTCHAs have shown broader acceptability due usability security factors. However, recent progress in deep learning implies that text-CAPTCHAs can easily be exposed fraudulent attacks. Thus, are getting research attention enhance security. In this work, the neural-style transfer (NST) adapted for designing an image-CAPTCHA algorithm while maintaining human performance. NST-rendered image-CAPTCHAs, existing methods inquire a user identify or localize salient object (e.g., content) which solvable effortlessly by off-the-shelf intelligent tools. Contrarily, we propose Style Matching CAPTCHA (SMC) asks select style image applied NST method. A solve random SMC challenge understanding semantic correlation between content output as cue. The performance solving evaluated based 1368 responses collected from 152 participants through web-application. average accuracy three sessions 95.61%; response time each per 6.52 s, respectively. Likewise, Smartphone Application (SMC-App) devised using proposed SMC-App 96.33%, 5.13 s. To evaluate vulnerability of SMC, learning-based attack schemes Convolutional Neural Networks (CNN), such ResNet-50 Inception-v3 simulated. considering studies 37%, improved over methods. Moreover, in-depth analysis, experimental insights, comparative imply suitability SMC.