Master of Engineering Science
Electrical and Computer Engineering
The field of authentication has a lot of room to develop in the age of big data and machine learning. Conventional high-accessibility authentication mechanisms including passwords or security questions struggle with critical vulnerabilities, creating a need for alternative authentication mechanisms able to cover said weaknesses.
We sought to create an authentication mechanism that creates dynamic, ever-changing security questions only the user can answer while remaining intuitive to use and as accessible as typical security questions by creating an authentication chatbot that leverages big data and natural language processing to pose dynamic authentication challenges.
We tested the components of our design in simulated scenarios to prove their efficacy, and found that all critical elements of the design can satisfactorily complete the tasks set for them. Thus we believe this design offers a useful supplement or alternative to password or security question-based authentication, improving the security of user data in our society.
Summary for Lay Audience
Passwords are vulnerable and insecure, and it's hard to fix that. Easily guessed passwords, password reuse, leaked databases, they're simple to use but carry not-insignificant risk.
Many modern systems improve security by layering multiple authentication mechanisms together, adding security questions or codes delivered via text message. Here we invent a new mechanism for this purpose, a chatbot designed to be convenient and straightforward to use yet secure and reliable.
The chatbot functions by taking unusual transactions or other recorded events about a user, ones that they're likely to remember, and automatically forming questions that only the real user can correctly answer. Artificial intelligence is extensively used in the process, allowing the system to get better and better the more people it authenticates.
Voege, Peter, "Smart Chatbot For User Authentication" (2021). Electronic Thesis and Dissertation Repository. 8315.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.