NeuralShield is an advanced, machine learning-powered spam detection system designed to protect users across multiple digital communication platforms. Initially launched as a desktop application, its core detection engine is being actively developed into a comprehensive suite of tools including a web API, a browser extension, and more.
In today's interconnected world, digital communication is essential. However, this connectivity comes with a cost: an relentless onslaught of spam and scams.
- Financial Loss: Phishing emails, fake lottery scams (e.g., "Magento" scams on Telegram), and SMS fraud lead to billions of dollars in losses for individuals and businesses annually.
- Security Risks: Spam messages are a primary vector for malware, ransomware, and identity theft, compromising personal and organizational security.
- Productivity Drain: Sorting through irrelevant and malicious content wastes countless hours of human productivity.
- Erosion of Trust: The constant barrage of scams erodes trust in digital communication channels.
Millions of internet users are affected daily, from individuals losing their savings to large corporations facing data breaches. Current solutions are often platform-specific, leaving users vulnerable across the digital landscape.
NeuralShield tackles this problem head-on with a unified, intelligent detection engine.
We have developed a highly accurate Logistic Regression model that is trained on a massive corpus of legitimate and spam content. This model excels at identifying the subtle linguistic patterns and markers common to scams across different platforms.
- Cross-Platform Protection: Currently detects spam in Emails, SMS, Telegram messages, and Instagram DMs. Our roadmap includes expansion to all major social media and messaging platforms (WhatsApp, Facebook, Twitter, etc.).
- Proactive Desktop GUI: The current desktop application allows for real-time analysis and classification of text content on your computer.
- High Accuracy & Performance: The Logistic Regression model provides an excellent balance of high prediction accuracy, interpretability, and low computational cost, making it ideal for real-time applications.
- Privacy-Focused: The core model can perform analysis locally on the user's device (in the desktop app and future browser extension), ensuring that sensitive messages never leave your machine.
This repository contains the core spam detection module and the flask web app.
| Component | Status | Lead | Description |
|---|---|---|---|
| Core Detection Module | ✅ Stable | Team | The trained Logistic Regression model and processing code. |
| Desktop GUI Application | 🔄 Active Development | Team Leader | Feature-complete standalone desktop app. |
| RESTful API | ✅ Stable | Team Leader | A web service to allow integration with other apps. |
| Browser Extension | 🔄 Active Development | Team Leader | Real-time protection for webmail and social media sites. |
| Public Website | ✅ Stable | Team Leader | Landing page with documentation and demo. |
- Python 3.8 or higher
- pip
-
Clone the repository:
git clone https://github.com/GreatTitanDev/NeuralShield-repo.git cd NeuralShield-repo -
Install dependencies:
pip install -r requirements.txt
-
Run the application:
python run.py
-
Use the web app: Paste text into the input box and click "Analyze" to get an instant spam classification.
We are building the future of spam protection and welcome contributions! Our focus is on expanding platform coverage and improving model accuracy.
Areas where we need help:
- Data Collection: Helping us gather and label spam/ham datasets from various platforms.
- Feature Engineering: Improving the text preprocessing and feature extraction.
- Frontend Development: Contributing to the website and browser extension (JavaScript/HTML).
- API Development: Helping build out the Flask/Django REST API.
Please read our CONTRIBUTING.md for guidelines on submitting pull requests.
This project is licensed under the MIT License - see the LICENSE file for details.
For questions, collaboration, or to report issues, please open an issue on this repository or contact the development team lead at Nimona Engida or Nimona Engida.
Disclaimer: This tool is designed to be an assistive layer of defense. Users should always remain vigilant and practice good digital hygiene, as no automated system is 100% foolproof.