A New Approach for Labeling Images Using CAPTCHA in Image Semantic Search Engines

Abstract

In this paper, we present PICTCHA as a CAPTCHA system (Completely Automated Public Turing test to tell Computers and Humans Apart). This system is a security tool for recognition of human users and instead of a complex text, it uses image labels as a valuable output. PICTCHA is made by Web images. In order to verify this CAPTCHA, users are obliged to enter two words for labeling (naming) a couple of images. In case of convenient names for the images, the presented meaningful names are used for specifying the content of images. Furthermore, meaningful graphics are created for names and images according which we may develop an image semantic search engine. Due to benefiting from images in the proposed system and its architecture, it has higher security level in comparison with other rivals. In experiments with 60 participants, the correctness of PICTCHA words was %98.18 while about %61.26 users verified this challenge successfully.