Intellectual Property Office

Non-Confidential Disclosures

“Image-based CAPTCHAs for Website Security”

PSU Invention Disclosure No. 3125
Download a PDF of this description

Field of the Invention/Key Words:

CAPTCHAs; Web security; Internet authentication or verification; automated Turing test, image retrieval

Links:

Inventor Website-1
Inventor Website-2
CAPTCHA

Inventors:

R. Datta, J. Wang and J. Li

Background:

Many Websites require an authentication process whereby a test is given to tell humans and computers apart to help prevent automated use of the Website by computers. When a computer program is able to generate such tests and evaluate the result, it is known as a CAPTCHA (Completely Automated Public test to Tell Computers and Humans Apart). In the past, Websites have often been attacked by malicious programs that register on a massive scale. Programs can be written to automatically consume large amounts of Web resources or bias results in on-line voting. This has driven many Websites to the idea of CAPTCHA-based security to ensure that such attacks are not possible without human intervention, which in turn makes them ineffective. A CAPTCHA acts as a security mechanism by requiring a correct answer to a question which only a human can answer any better than a random guess. Most current CAPTCHAs are text-based and an example is shown to the right (Fig 2). Commercial CAPTCHAs play a significant role in enhancing the security and reliability of the World Wide Web. Some of these can be found while registering for a new Yahoo! Account or signing up for
PayPal. However, text-based CAPTCHAs have been broken using object-recognition techniques with high accuracies. This reduces the reliability of security protocols based on text based CAPTCHAs.

Invention description:

We have created a system for the generation of attack-resistant,
user-friendly, image-based CAPTCHAs. We produce controlled distortions on randomly chosen images and present them to the user for annotation from a given list of words (Fig 1). The images are distorted in a way that prevents the chance of successfully using image understanding technologies such as content-based image retrieval. In a preferred implementation of our technology this selection is combined with the prior step of clicking near the center of any picture in a mosaic such as that shown to the right (Fig 3). This results in a two-round click-and-annotate process that makes the process both user friendly and very effective.

Contact:

Bradley A. Swope
Sr. Technology Licensing Officer
The Pennsylvania State University
113 Technology Center
University Park, PA 16802
Phone: (814) 863-5987
Fax: (814) 865-3591
E-mail: bradswope@psu.edu