PhishSense

Voiceover

Presenter Information(s)

Ashlynn Hainey, Deveyn Hainey

Abstract or Description

94% of organizations were victims of phishing attacks in 2024, and 96% of those were negatively impacted. PhishSense is an artificial intelligence-powered Gmail add-on designed to help users identify, understand, and educate themselves on potential phishing threats all in their inbox. The goal of this project is to bridge the gap between identifying and educating users. In order to do so, PhishSense enables users by allowing them to choose what emails they wish to scan, which PhishSense securely extracts metadata, stores relevant information in a hosted database, and then leverages large language models to classify the message as “Safe” or “Phishing”. Utilizing this data, the project will provide the user with statistics and relevant information to make an informed decision regarding the email. The testing methodology includes benchmarking PhishSense’s accuracy against known phishing datasets, assessing false positive/negative rates, and analyzing user interactions to measure effectiveness. Through preliminary testing, PhishSense achieved 88% accuracy, surpassing the average human detection rate of about 80%. PhishSense helps bridge this gap by not only identifying threats but also educating users on how to recognize and avoid them. By combining large language model technology, cloud-based big data services, and a user-friendly UI, PhishSense aims to reduce the likelihood of personal loss, deception, and other cyber risks associated with email-based attacks. Unlike traditional phishing detection tools that solely flag threats, PhishSense is built to educate users—helping them understand why an email is phishing, what red flags to look for, and how to make safer decisions in the future.

Mentor

Dr. Zhengxiong Li