Sentry - Phishing and Spam Detection Website
A web application for checking suspicious links and messages, with risk classification, clear explanations, and limited guest access before registration.
Sentry is a web application designed to help users assess suspicious URLs and messages before clicking links, responding to senders, or sharing sensitive information. Built with React, Vite, and Tailwind CSS, the frontend integrates with a FastAPI backend to request analysis and present the results in an accessible interface.
Users can try detection without an account, subject to a backend-managed usage limit. Once the limit is reached, they are prompted to log in or register without losing their current input. Account features include email verification, password recovery, session renewal, and personal scan history.
The interface combines a dark visual theme, an animated radar, responsive navigation, and educational content. Results include a risk classification, model confidence where available, and recommended next steps. A screenshot preview and local simulation demonstrate the planned image-analysis experience.
Sentry supports informed decision-making; its results are guidance, not a guarantee that content is safe.
My contribution
Engineering decisions
Making phishing and spam analysis understandable to non-technical users while integrating separate frontend and backend services. The project also needed to balance convenient guest access with usage limits, handle authentication and API failures, and clearly distinguish real detection results from simulations. Finding datasets for training ML models is also a challenge for me because it's limited and hard to find, because of that for the screenshot detection feature is still unavailable for use.
Developed a guided scanning interface with separate URL and message inputs, readable result summaries, and actionable recommendations. Implemented backend-driven guest limits, login and registration prompts that preserve input, session renewal, and API error handling. Responsive layouts and clearly labeled simulations help users understand both the workflow and the system’s limitations.