Date on Master's Thesis/Doctoral Dissertation
Computer Engineering and Computer Science
Computer Science and Engineering, PhD
Yampolskiy, Roman Vladimirovich
Committee Co-Chair (if applicable)
Imam, Ibrahim N.
Hardin, Charles Timothy Hardin
Human face recognition (Computer science); Computer vision; Computer security; Machine learning
Bots are malicious, automated computer programs that execute malicious scripts and predefined functions on an affected computer. They pose cybersecurity threats and are one of the most sophisticated and common types of cybercrime tools today. They spread viruses, generate spam, steal personal sensitive information, rig online polls and commit other types of online crime and fraud. They sneak into unprotected systems through the Internet by seeking vulnerable entry points. They access the system’s resources like a human user does. Now the question arises how do we counter this? How do we prevent bots and on the other hand allow human users to access the system resources? One solution is by designing a CAPTCHA (Completely Automated Public Turing Tests to tell Computers and Humans Apart), a program that can generate and grade tests that most humans can pass but computers cannot. It is used as a tool to distinguish humans from malicious bots. They are a class of Human Interactive Proofs (HIPs) meant to be easily solvable by humans and economically infeasible for computers. Text CAPTCHAs are very popular and commonly used. For each challenge, they generate a sequence of alphabets by distorting standard fonts, requesting users to identify them and type them out. However, they are vulnerable to character segmentation attacks by bots, English language dependent and are increasingly becoming too complex for people to solve. A solution to this is to design Image CAPTCHAs that use images instead of text and require users to identify certain images to solve the challenges. They are user-friendly and convenient for human users and a much more challenging problem for bots to solve. In today’s Internet world the role of user profiling or user identification has gained a lot of significance. Identity thefts, etc. can be prevented by providing authorized access to resources. To achieve timely response to a security breach frequent user verification is needed. However, this process must be passive, transparent and non-obtrusive. In order for such a system to be practical it must be accurate, efficient and difficult to forge. Behavioral biometric systems are usually less prominent however, they provide numerous and significant advantages over traditional biometric systems. Collection of behavior data is non-obtrusive and cost-effective as it requires no special hardware. While these systems are not unique enough to provide reliable human identification, they have shown to be highly accurate in identity verification. In accomplishing everyday tasks, human beings use different styles, strategies, apply unique skills and knowledge, etc. These define the behavioral traits of the user. Behavioral biometrics attempts to quantify these traits to profile users and establish their identity. Human computer interaction (HCI)-based biometrics comprise of interaction strategies and styles between a human and a computer. These unique user traits are quantified to build profiles for identification. A specific category of HCI-based biometrics is based on recording human interactions with mouse as the input device and is known as Mouse Dynamics. By monitoring the mouse usage activities produced by a user during interaction with the GUI, a unique profile can be created for that user that can help identify him/her. Mouse-based verification approaches do not record sensitive user credentials like usernames and passwords. Thus, they avoid privacy issues. An image CAPTCHA is proposed that incorporates Mouse Dynamics to help fortify it. It displays random images obtained from Yahoo’s Flickr. To solve the challenge the user must identify and select a certain class of images. Two theme-based challenges have been designed. They are Avatar CAPTCHA and Zoo CAPTCHA. The former displays human and avatar faces whereas the latter displays different animal species. In addition to the dynamically selected images, while attempting to solve the CAPTCHA, the way each user interacts with the mouse i.e. mouse clicks, mouse movements, mouse cursor screen co-ordinates, etc. are recorded nonobtrusively at regular time intervals. These recorded mouse movements constitute the Mouse Dynamics Signature (MDS) of the user. This MDS provides an additional secure technique to segregate humans from bots. The security of the CAPTCHA is tested by an adversary executing a mouse bot attempting to solve the CAPTCHA challenges.
D’Souza, Darryl Felix, "Avatar captcha : telling computers and humans apart via face classification and mouse dynamics." (2014). Electronic Theses and Dissertations. Paper 1715.