Fraud Shield Web

Inspiration

E-commerce giants lose billions of dollars annually due to return fraud—where customers return fake, damaged, used, or completely wrong products instead of the original items. Traditional systems process returns only after the product arrives back at the warehouse. By then, logistics and shipping costs are already wasted. Inspired by this massive industry gap, we wanted to build a real-time validation mechanism right at the customer's doorstep before logistics costs are incurred.

What it does

FraudShield Web is an AI-powered e-commerce return verification platform that operates through two primary modules:

Customer Return Portal: Requires customers to initiate return requests by taking real-time photos of the product and scanning the serial number/barcode via their device camera.

Admin Fraud Inspection Panel: Processes the captured images through Computer Vision models to verify authenticity and calculates an instant Fraud Risk Score (%). If high risk is detected, the request is flagged or rejected before pickup.

How we built it

Frontend: Developed using React.js / HTML5 with WebRTC APIs for seamless real-time camera streaming and image capture.

Styling & UI: Customized Bootstrap and CSS3 for a responsive, clean dashboard interface.

Backend: Powered by Node.js / Express (or Python FastAPI) to handle return requests and process API workflows.

AI & Computer Vision: Integrated OpenCV / Google Teachable Machine APIs for image verification and matching original product parameters against returned items.

Challenges we ran into

Real-time Camera Streaming: Ensuring smooth camera streaming and barcode scanning directly in web browsers across different mobile and desktop devices.

Image Quality Variations: Handling lighting differences, blurry photos, and low-resolution camera inputs while maintaining accurate AI Fraud Risk Score calculations.

Low-Latency Analysis: Optimizing backend response time so the verification process completes in seconds without delaying the user interface.

Accomplishments that we're proud of

Successfully building an end-to-end working prototype that connects live web camera input with a functional fraud detection algorithm.

Designing an intuitive Admin Panel that instantly visualizes Fraud Risk Scores with actionable flags (Approve / Review / Block).

Creating a solution that addresses both user convenience and corporate loss prevention effectively.

What we learned

Hands-on experience integrating Computer Vision and AI models into modern web applications using web APIs.

How e-commerce logistics and supply chain reverse workflows operate, and where automated tech intervention yields maximum ROI.

Best practices in frontend state management and building responsive dashboards under tight hackathon deadlines.

What's next for Fraud Shield Web

AWS Rekognition Integration: Upgrading our vision pipeline to enterprise AWS Cloud services for higher accuracy and large-scale image processing.

Logistics Boy App Integration: Extending the web portal into a mobile PWA for delivery agents to scan and verify products physically at the doorstep during pickup.

Predictive User Behavioral Analytics: Building ML models that analyze customer purchase and return history to flag serial refund abusers automatically.

Built With

• bootstrap
• express.js
• google-teachable-machine-api
• html5-(webrtc)
• javascript
• node.js
• opencv
• react.js