BankGuard AI

BankGuard AI was inspired by the growing need for seamless, real-time security in mobile banking. Traditional methods like OTPs and biometrics can be spoofed or bypassed, so we built a system that continuously authenticates users based on behavioral biometrics—like typing speed, touch pressure, and swipe patterns—using lightweight ML models running directly on-device. The backend, built with Node.js and Supabase, logs anonymized session data and powers a React-based admin dashboard to visualize fraud risk. We faced challenges in ensuring high model accuracy without draining device performance, while staying compliant with GDPR and DPDP. We're proud of creating a privacy-first, user-friendly fraud detection system, and plan to scale it through pilots with fintechs, add Retrieval-Augmented Generation (RAG) for context-aware detection, and open-source its core behavioral engine for ethical AI security solutions.