🚨 Can AI Stop Online Payment Fraud? | Real-Time Fraud Detection System 🤖

Описание: Can AI make online payments safer without adding unnecessary friction? 🔐

In this BIOS Hackathon 2026 project, *Team nexus_4* presents a Fast and Secure Online Transaction Authentication system designed to assess transaction risk and make smarter authentication decisions.

🚀 *How does the project work?*

Our proposed system combines rule-based checks with *LightGBM machine learning* to evaluate transaction risk and choose one of three outcomes:

✅ *ALLOW* — for low-risk transactions
⚠️ *CHALLENGE* — when additional verification may be needed
🚫 *BLOCK* — for high-risk transactions

🔍 *Key technologies and concepts*
FastAPI for transaction-risk API processing
Rule-based checks and LightGBM risk scoring
Transaction signals such as amount, device status and recent activity
SHAP and rule hits for explainable decisions
A dashboard to help analysts understand why a transaction was challenged

🎯 *Our goal:* Make digital payments safer while reducing unnecessary authentication friction for genuine users.

📊 *Important note:* The risk scores shown in the video are illustrative examples, device and behavioural signals are simulated, and the sub-200-millisecond latency figure is a target—not a measured result.

🏆 *BIOS Hackathon 2026*
👥 *Team:* nexus_4
🎓 *Institute:* Thapar Institute of Engineering & Technology, Patiala

Team members: Shreya Sharma, Shubhi Singh, Subham Kumar and Tanush Grover.

💬 What do you think matters most in online payment security: preventing fraud, reducing false alarms, or making authentication faster? Share your thoughts in the comments.

#ArtificialIntelligence #FraudDetection #Hackathon