Random Handwritten CAPTCHA: Web Security with a Difference

Abstract

Bots are automated programs designed to make auto registrations in online services, resulting in wastage of resources and breach of web security. English based CAPTCHAs are used to prevent bots from abusing these online services. However, English based CAPTCHAs have some inherent flaws and have been broken by bots. In this paper, an Arabic text based CAPTCHA is proposed. The CAPTCHA text image is distorted with background noise. Background noise and dots in the Arabic text makes CAPTCHA hard to be broken by Arabic OCRs. The proposed scheme is useful in Arabic speaking countries and in protecting internet resources. The proposed CAPTCHA scheme is both secure and robust. Experimental results show that background noise is a good defense mechanism against OCR recognizing Arabic text. © 2010 Springer-Verlag.