TinyPhish
Project Overview
An AI-powered phishing detection platform developed for the TinyFish Hackathon to combat AI-driven scams and social engineering attacks.
Scans suspicious URLs within a sandboxed browser, protecting the user's device from malware while extracting live content for analysis.
Leverages Gemini to detect phishing patterns (e.g., homograph attacks) and integrates with VirusTotal to query over 70 security vendors for domain reputation.
Tech Stack
Claude Opus 4.6Gemini APIReactTypeScriptViteTailwind CSSFastAPIPythonSupabasePostgreSQLVirusTotal API
My Contributions
Developed a React frontend using Vite, Tailwind CSS, and shadcn/ui to stream live browser scan results via SSE (Server-Sent Events) from the TinyFish Agent API.
Built a Python backend using FastAPI and Pydantic to orchestrate parallel API calls to OpenAI for content analysis and VirusTotal for threat intelligence.
Designed and implemented the PostgreSQL database schema on Supabase to persist scan results, enabling instant history retrieval.
Implemented an intelligent caching system with a smart Time-To-Live (TTL) to instantly deliver results for known sites, reducing API costs and staying within rate limits.
Learning point : Utilizing Server-Sent Events (SSE) allows for efficient, real-time streaming of browser automation data to the frontend without the overhead of WebSockets.
Project Links