Under the direction of Dr. Tom Goldstein in UMD’s machine learning capstone course, I am working on a project designed to test the effectiveness of large language models for hyper-targeted phishing attacks by creating and deploying phishing emails to UMD Computer Science faculty.
I was the team member in charge of email generation and developed all of the language model prompts that were used to create the emails; these prompts directed the model to extract information from faculty biographies, synthesize emails from that information, and automate selection of the best email to send.