An efficient technique for breaking of coloured Hindi CAPTCHA

Cyber-attacks on the internet are not a new phenomenon. Hackers are always looking for ways to breach the security of the internet. Good researchers continuously develop solutions to the problems of breaches in security. Human interaction proofs are also one of the most affordable solutions to problems. The completely automated public Turing test to tell computers and humans apart (CAPTCHA) has a wide range of applications, including website authentication, preventing bogus email registration, preventing bogus online polls, preventing bogus ticket booking, and so on. CAPTCHAs come in various formats, but text-based schemes are the most common. Unfortunately, these text-based CAPTCHAs are available in foreign languages and have a high success rate. The usability of these foreign language-based schemes for native people is also very low. These days, Hindi CAPTCHAs are available, but before using them, we must test the security of these schemes. In this study, ten different coloured CAPTCHAs are broken using the k-nearest neighbours, the support vector machine, and the random forest classifiers. The breaking rate for two-colour schemes is 90%, while the breaking rate for multi-colour schemes is 93%. The solving time for the support vector machine and the random forest classifier ranges from 0.06 s to 0.09 s and 0.1 to 0.2 s in the case of two-colour schemes. For multi-colour schemes, it ranges from 0.06 s to 0.09 s and 0.09 s to 0.2 s for the support vector machine and random forest classifier, respectively.