Securing the Web Using Human Perception and Visual Object Interpretation

Abstract

Being motivated by the need for usable Web security systems, we present in this paper new image-based CAPTCHA systems that overcome the weaknesses of commercial CAPTCHAs. We describe Tree and Shape CAPTCHAs in addition to handwritten CAPTCHAs that feature visual objects transformed according to specific principles of cognitive psychology. These transformations, in addition to other security measures we describe, ensure that our CAPTCHAs are easily interpretable by humans but not by machines. We seek not only to provide useful CAPTCHAs, but to offer important insights into such fields as Information Visualization, Human-Computer Interaction, etc. Early testing results indicate that users find our CAPTCHAs an attractive alternative to text-based CAPTCHAs and are readily able to solve them based on cognitive abilities. Conversely, machine recognition is low due to the current inability to make use of these cognitive aids, thus making our CAPTCHAs a viable solution for cyber security.