Talk: A Privacy and Security Analysis of myUMBC Answers, 3/28

Generative AI has the potential to improve user search experiences by supporting natural language querying and providing more detailed and domain-specific responses. UMBC is seeking to provide this convenience to myUMBC users through the myUMBC Answers system, which enables users to access both personal and UMBC services information from the myUMBC search bar. In this talk, we investigate the resiliency of Answers against several common generative AI attacks that were performed by the UMBC Scholarship for Service (SFS) scholars in collaboration with other students and the UMBC Division of Information Technology (DoIT). We first provide an overview of the study and the myUMBC Answers system before discussing the types of attacks which were launched against the system. We then explore the behavior of the Answers system in response to these attacks. Finally, we outline the recommendations provided to DoIT by the study participants to improve the security and user experience of myUMBC Answers.

Christian Badolato is a PhD student working with Professor Roberto Yus focusing on data privacy in the Internet of Things at UMBC after having received his master’s degree from the same university. He has several years of experience as a software architect and is a Certified Information Systems Security Professional.

Support for this UMBC Cyber Defense Lab event was provided in part by NSF under SFS grant DGE-1753681.