About Us

Overview

IMAGINATION, initially conceived in 2005 and launched as a demo system in April 2008, is an image-based CAPTCHA system that uses image recognition as a basis for distinguishing humans and machines. CAPTCHAs (Completely Automated Public Turing test to tell Computers and Humans Apart), also called HIPs (Human Interactive Proof), by doing this distinction, help reduce e-mail spam, stop automated blog and forum responses, save resources, and and prevent denial-of-service attacks on Web servers, among others. Text-based CAPTCHAs, such as those found here, are in widespread use commercially and otherwise.

Novelty

The problem with text-based CAPTCHAs is that there has been sizable number of successful attempts at breaking them, hence making security mechanisms based on these vulnerable to attacks. Some of these success stories have been documented in research publications, particularly in the computer vision and OCR research community, with success rates up to 90%. Following is a list of some of these publications.

1. K. Chellapilla and P. Y. Simard, "Using Machine Learning to Break Visual Human Interaction Proofs (HIPs)," Proc. NIPS, 2004.
2. G. Moy, N. Jones, C. Harkless, and R. Potter, "Distortion Estimation Techniques in Solving Visual CAPTCHAs," Proc. IEEE CVPR, 2004.
3. G. Mori and J. Malik, "Recognizing Objects in Adversarial Clutter: Breaking a Visual CAPTCHA," Proc. CVPR, 2003.
4. A. Thayananthan, B. Stenger, P. H. S. Torr, and R. Cipolla, "Shape Context and Chamfer Matching in Cluttered Scenes," Proc. IEEE CVPR, 2003.

Therein lies the novelty of our IMAGINATION System, which requires solving a harder AI problem, that of image recognition, in order to break. Therefore, in principle, the system is more secure than text-based CAPTCHAs, with image recognition being a harder problem, and the "space" of images being much larger. Its design, which includes application of careful, randomly generated distortions on the fly, has been based on its robustness to attacks by image recognition/retrieval algorithms including our group's SIMPLIcity system. This is not to say that breaking of IMAGINATION is not possible with other algorithms or approaches, or some combination of smarts. We invite other researchers to attempt to break IMAGINATION, which would (a) help us improve the system, and (b) push the state-of-the-art in image recognition under heavy distortion. We believe image-based CAPTCHAs are a viable alternative to text-based ones in overcoming some of their limitations.

Personnel

IMAGINATION has been mainly developed by Dr. Ritendra Datta when he was a doctoral student under the guidance of Prof. James Z. Wang and Prof. Jia Li. He is with Google currently. User studies on IMAGINATION were conceived, developed, and conducted by Dr. Dhiraj Joshi during his graduate study at Penn State. He is now a Scientist at Kodak Research, Rochester, NY. Razvan Orendovici, and undergraduate honors student in IST/CSE at Penn State, has designed and developed much of the demonstration front-end, and concurrency control.

Academic Publications

1. Original Article on the IMAGINATION System: R. Datta, J. Li, and J.Z. Wang, "IMAGINATION: A Robust Image-based CAPTCHA Generation System," Proceedings of the ACM Multimedia Conference, pp. 331-334, Singapore, ACM, November 2005. [PDF]


2. Journal Article: R. Datta, J. Li, and J.Z. Wang, "Exploiting the Human-Machine Gap in Image Recognition for Designing CAPTCHAs," IEEE Transactions on Information Forensics and Security, vol. 4, no. 3, pp. 504-518, 2009. [PDF]

Ongoing Development

For information on the current development of this CAPTCHA, please check out this new Website.

Sponsorship

This material is based upon work supported by the National Science Foundation under Grant No. 0347148. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

If you have questions or comments, please email James Wang at .