ScamShield AI: Detect Scams Locally with Open-Source AI 🛡️
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.
What I Built
I built ScamShield AI, an AI-powered scam detection tool designed to help people identify suspicious messages, phishing attempts, fraudulent links, and potential online scams.
Online scams are becoming increasingly sophisticated, and many people struggle to distinguish legitimate messages from fraudulent ones. ScamShield AI aims to make scam detection easier by analyzing user-provided text and explaining the warning signs in simple language.
Key Features
- 🛡️ AI-Powered Analysis: Analyzes suspicious messages using a locally running AI model.
- 🔍 Scam Detection: Identifies potential phishing, fraud, impersonation, and social engineering attempts.
- 💻 Local AI Inference: Uses Ollama to run an open-weight language model locally.
- 🔐 Privacy-Focused Design: Message analysis can run locally without sending message contents to a third-party AI API.
- 📋 Explainable Results: Provides understandable reasons why a message may be suspicious.
- 🌐 Simple Web Interface: Makes scam analysis accessible through a user-friendly interface.
Who Could Benefit?
ScamShield AI could be useful for students, individual internet users, freelancers, small businesses, and organizations that want an additional layer of protection against online scams.
The project could eventually evolve into a freemium security tool, a browser extension, or a subscription-based service for small businesses.
ScamShield AI is a decision-support tool, not a guarantee that a message is safe or fraudulent.
Demo
Live Demo: https://vedantshukla-cracker.github.io/ScamShield-AI/
The demo will showcase how users can submit a suspicious message, analyze it with AI, and review the generated risk assessment and explanation.
Code
GitHub Repository: https://github.com/vedantshukla-cracker/ScamShield-AI
The project source code, installation instructions, configuration, and testing guidance will be available in the repository.
Contributions, bug reports, feature requests, and improvements are welcome!
How I Built It
ScamShield AI combines a lightweight web interface with local AI inference to analyze potentially fraudulent messages.
Technology Stack
- HTML, CSS, and JavaScript: Frontend interface and user interactions.
- Ollama: Runs the AI model locally.
- Open-Weight Language Model: An instruction-following model suitable for text classification and scam analysis.
- Python / Backend (if applicable): Connects the web interface to the local inference service.
How It Works
- The user enters a suspicious message.
- The application sends the text to the locally running model through Ollama.
- The model analyzes possible scam indicators, including urgency, suspicious requests, impersonation, and requests for sensitive information.
- ScamShield AI displays a risk assessment and an explanation of the warning signs.
- The user can make a more informed decision about the message.
Open-Source AI and Model Licensing
The project is designed around local inference using Ollama and an open-weight model.
The exact model name, version, and license should be documented in the repository before release. Model licenses differ, so the selected model's license must be checked independently rather than assuming every Ollama model is fully open-source.
Local inference reduces dependence on paid external AI APIs and can improve privacy by keeping message content on the user's machine, provided the application does not transmit that content through other services.
Limitations and Testing
ScamShield AI is an experimental security assistant. It may produce false positives or false negatives, and its results should not replace independent verification.
Testing should include:
- Legitimate messages to check for false alarms.
- Known phishing and scam examples.
- Messages containing urgent payment requests.
- Impersonation and credential-stealing attempts.
- Empty, malformed, and unusually long inputs.
- Verification that message contents are not unintentionally transmitted to external services.
The goal is to build a useful, privacy-conscious starting point for accessible AI-assisted scam awareness.
Why Does Open Innovation Matter?
Open innovation makes projects like ScamShield AI more accessible, transparent, and adaptable.
By using an open-weight model and local inference, developers can experiment with AI without depending entirely on proprietary APIs or paying for every inference request.
It also creates opportunities to:
- Inspect and improve the application's source code.
- Experiment with different models and compare their performance.
- Adapt the tool for different languages and scam patterns.
- Encourage community contributions to testing and security.
- Learn about practical AI deployment, privacy, and responsible development.
Open innovation does not automatically guarantee security or unrestricted model usage. Model licenses, dependencies, and data-handling practices still need to be reviewed carefully.
For me, this project is an opportunity to explore how open AI can solve a real-world problem while making AI development more approachable.
My Agent Session
[Optional: Add your DevRelay agent session link here.]
This section can document how an AI coding agent helped with implementation, testing, debugging, or project development.
Prize Categories
Challenge: Hacktoberfest Open-Source AI Challenge — Week 2: Side Hustle.
Project Focus: Open-source AI, local inference, scam awareness, and privacy-conscious security tooling.
[Add the official partner prize categories you are entering, after checking the challenge rules.]
Built with curiosity, open-source AI, and a mission to make the internet a little safer. 🛡️
Top comments (0)