On Random Field CAPTCHA Generation
Date
Author
Institution
Degree Level
Degree
Department
Specialization
Supervisor / Co-Supervisor and Their Department(s)
Examining Committee Member(s) and Their Department(s)
Citation for Previous Publication
Link to Related Item
Abstract
In this thesis, we develop a novel method of generating CAPTCHAs, which are used to protect online resources from abuse by computer agents. We view CAPTCHA generation as random field simulation and construct a CAPTCHA by evolving an initial state via resimulating pixels until the image becomes readable. We empirically demonstrate that this CAPTCHA is easy for humans to read but difficult for computer programs to crack.
We describe how to develop variants of this CAPTCHA; in particular, we implement and assess the utility of a grey-level variant. We establish a method of maximizing the effectiveness of a CAPTCHA variant, and perform analysis to determine which properties of the CAPTCHA most effectively differentiate humans and computer programs.
We extend the random field used in the CAPTCHA application to multiple dimensions in the context of graph theory, and describe the generic method of applying the random field to suitable problems.