Face Recognition CAPTCHAs

Web-based services are becoming more and more ubiquitous and are replacing human-to-human interactions. Automated programs abuse the web services by pretending to be human. As a result, the need to authenticate that the other party on the web is a human and not a program, is on the rise. CAPTCHA is a test that can be used to reliably differentiate between human users and automated programs on the web. In this paper, we propose a new CAPTCHA scheme based on the problem of face recognition. This test takes advantage of the fact that recognizing human faces is considered to be a tough task for computers, but is relatively easy for humans. In the proposed test, human face photographs from a public database are distorted using two different image processing transformations. The user is asked to match distorted photographs of several different human subjects. The automatic generation and evaluation of tests is shown to be possible using the image processing open-source tool Gimp. The proposed CAPTCHAs have the desirable properties of being easy for humans while being difficult for programs to solve. Also the level of comfort in passing these tests is high, independently of the person’s familiarity with the English language, when compared to other English text based CAPTCHAs.