Browse Prior Art Database

CAPTCHA display using a moving viewpoint that never fully shows an object in any frame

IP.com Disclosure Number: IPCOM000222373D
Publication Date: 2012-Sep-27
Document File: 4 page(s) / 137K

Publishing Venue

The IP.com Prior Art Database

Abstract

CAPTCHAs (Completely Automated Public Turing tests to tell Computers and Humans Apart) are typically used to protect publically accessible web resources from unwanted access. This article seeks to increase the difficulty of automated algorithmic solvers whilst at the same time making it easy for humans to solve the CAPTCHA problem. A method of utilising a human's innate ability for assembling multiple incomplete views of imagery to recognise a single entity is presented.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 29% of the total text.

Page 01 of 4

CAPTCHA display using a moving viewpoint that never fully shows an object in any frame

The term "CAPTCHA" is an acronym based on the word capture and stands for "Completely Automated Public Turing test to tell Computers and Humans Apart". CAPTCHAs are typically used to protect publically accessible web resources from unwanted access by automated systems or low-cost human "farms". They are automatically generated and they present a new problem on each access attempt. The resources that CAPTCHAs protect must be accessible to diverse audiences - including non-native speakers of the language used and also disabled users - and so the problems must be conceptually and practically simple. A different mode of delivery of the task should be made available for disabled users or for those who otherwise find the visual task too difficult (for example, using an audio problem rather than a visual one).

    A basic task required by a typical visual CAPTCHA is the correct annotation of the characters in a word. For example, the CAPTCHA2 system [1] geometrically distorts two words in a noisy 2D image; furthermore, the font, distortions, position and colour of the foreground and background can be varied. The image is typically static, but some CAPTCHAs are animated and incorporate random motion of the characters to ensure some overlap in each frame, which increases the difficulty and requires a number of frames to be analysed [6]. Further developments include 3D rendering and complex backgrounds used in NuCAPTCHAII [2], exploiting visual illusions to defeat automated solvers [7]. and hiding characters behind moving objects [8].

    Ongoing improvements in computing power and the accuracy of computer vision algorithms, alongside improved training and other methods for the human "farms", mean that CAPTCHAs are being solved more quickly and reliably, leading to the erosion of the protection afforded by CAPTCHAs [3]. Therefore, more costly CAPTCHAs which aim to increase the difficulty of the problem or effectively increase the dictionary size [4], or prevent excessive numbers of accesses by the same attacker or increase the bandwidth required to solve the problem [5], are being sought. In this disclosure we aim to reduce the rate of unwanted accesses to resources protected by CAPTCHAs, while at the same remaining accessible and not inducing frustration in real users. With the increasing requirement for websites to protect their resources, CAPTCHAs represent a large potential revenue stream for companies that licence their technology for use by website owners [9].

    This disclosure aims to make automatic video analysis of a CAPTCHA more difficult, while exploiting both the increasing prevalence of high-speed consumer internet connections as well as the innate ability of the visual system of humans to assemble multiple incomplete views of imagery taken from different perspectives. The core innovation is the introduction of a viewpoint that jumps around the scene...