Distortion estimation techniques in solving visual CAPTCHAs

Abstract

Abstract-This paper describes two distortion estimation techniques for object recognition that solve EZ-Gimpy and Gimpy-r, two of the visual CAPTCHAs ("Completely Automated Public Turing test to tell Computers and Humans Apart") with high degrees of success. A CAPTCHA is a program that generates and grades tests that most humans can pass but current computer programs cannot pass. We have developed a correlation algorithm that correctly identifies the word in an EZ-Gimpy challenge image 99% of the time and a direct distortion estimation algorithm that correctly identifies the four letters in a Gimpy-r challenge image 78% of the time.

CAPTCHA: A Security Review

“…Relative shape context descriptors compared with the shape context descriptors in [13] has some obvious advantages: (1).not need the special selected polar coordinates of each point in the positive direction to reach the full rotation invariance, hence, has translational and rotational invariance; (2).not necessary to calculate items logρ,for all the points just only to be considered its positive direction of the polar coordinate polar angle, so possess the scale invariance; (3).not necessary to calculate the average distance from set point to all point to realize the regulation of distance entry, therefore, it has special point and point noise robustness. Pattern matching on [13] was based on a measure method of measuring similar to the shape.…”

Against Spyware Using CAPTCHA in Graphical Password Scheme

“…The security of our scheme relies on the robustness of CAPTCHA in resisting automatic adversarial attacks. However, it is not clear whether there is a true CAPTCHA at all and some reports show that some text-based CAPTCHAs can be partly or almost broken by automatic programs [3,29,30]. With the assumption that spyware is capable of detecting and recording screen snapshots, entered strings and the system feedback, we will analyze the security of the improved scheme from two extreme aspects.…”

iCAPTCHA: The Next Generation of CAPTCHA Designed to Defend against 3rd Party Human Attacks

“…First, background noise and anomalies are removed from a CAPTCHA image [2,3]. Next, the image is segmented and passed to character and shape recognition algorithms [3]. Utilizing this methodology, Jitendra Malik and Greg Mori were able to defeat the E-Z Gimpy CAPTCHA with a 92% success rate [2].…”

“…The CAPTCHA image could be more distorted. iCAPTCHA can be integrated with any current text-based CAPTCHA with feature such as juxtaposition [3]. In addition, the candidate button images must be dynamically generated.…”

“…Using image processing to attack a CAPTCHA is a multi-step process. First, background noise and anomalies are removed from a CAPTCHA image [2,3]. Next, the image is segmented and passed to character and shape recognition algorithms [3].…”