ScamShield India：利用本地 AI 協助識別網上騙案

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

I built ScamShield India, an AI-powered scam detection web application designed to help people identify suspicious messages and potential online scams.

Online scams are becoming increasingly sophisticated. Fraudulent messages can impersonate banks, delivery services, customer support teams, or government organisations. Many people struggle to distinguish legitimate communication from potential fraud.

ScamShield India aims to make scam awareness more accessible by allowing users to submit a message for analysis and receive an assessment of potential warning signs.

The project combines rule-based detection with a locally running AI model to analyse suspicious content.

The potential users include individuals, students, small businesses, and community organisations that want an accessible way to assess suspicious messages.

As a potential side hustle, ScamShield India could eventually become a subscription-based security awareness tool for small businesses or an affordable scam-screening service. It is currently a prototype, not an established commercial product.

Demo

GitHub repository: ScamShield India

The application currently runs locally on my laptop. I plan to demonstrate it by opening the website, submitting a fictional suspicious message, and showing the analysis results.

Demo video: [https://drive.google.com/file/d/1MU3iCMkgs38NED05SqPEIGUMqj3MTvI5/view?usp=sharing
]

Code

Explore the project on GitHub:

https://github.com/parvcore149/ScamShield_India

The project uses Python for its backend, a web interface for user interaction, and supporting modules for scam detection and AI integration.

How I Built It

I built ScamShield India using Python, FastAPI, HTML, and open-weight AI.

The main components include:

FastAPI: Provides the backend for the web application.
HTML: Provides the user-facing interface.
Rule-based detection: Identifies predefined patterns and warning signs associated with suspicious messages.
Gemma 3 through Ollama: Enables local inference using an open-weight AI model.
Python modules: Organise the detection pipeline and AI integration.

One of the key aspects of this project is experimenting with AI inference locally instead of relying entirely on a hosted AI API.

The application is a prototype, and its results should be treated as assessments rather than definitive proof that a message is safe or fraudulent.

Why Does Open Innovation Matter?

Open innovation made it possible for me to experiment with AI without building an entire model from scratch or depending exclusively on a paid, hosted AI service.

Using Gemma through Ollama gave me a way to explore local inference and build an application around an open-weight model on my own computer.

This approach offers several potential advantages:

Accessibility: Developers and students can experiment with AI using locally available tools.
Control: Developers can choose their model and customise their application architecture.
Privacy potential: Message analysis can remain on the user's machine when the complete processing pipeline runs locally and no external services are involved.
Learning and collaboration: An open code repository allows other developers to inspect the implementation, identify limitations, and suggest improvements.

Open innovation is especially valuable in cybersecurity because accessible tools can help more people experiment with ways to recognise and understand online threats.

My goal is to continue improving ScamShield India while learning more about AI, application security, and open-source development.

Prize Categories

Best use of GEMMA