TruthNet: AI-Powered Fake News Detection

TruthNet is a comprehensive, production-grade machine learning framework designed to identify and classify news articles as Real or Fake. By leveraging both classical statistical methods and state-of-the-art transformer architectures, TruthNet provides high-accuracy predictions accompanied by explainability signals to help users understand the reasoning behind every classification.

---

Table of Contents

1. Overview
2. Key Features
3. Architecture
4. Performance Benchmarks
5. Installation
6. Configuration
7. Training Pipeline
8. API Documentation
9. Frontend Interface
10. Contributing
11. License

---

Overview

In an era of rapid information dissemination, the proliferation of misinformation poses a significant challenge to public discourse. TruthNet addresses this by providing a robust, scalable API and web interface for news verification. It processes raw article text through a sophisticated preprocessing pipeline and classifies it using a suite of trained models, ranging from efficient Logistic Regression to deep-learning DistilBERT transformers.

Disclaimer: TruthNet is intended for research and educational purposes. No automated classifier is a substitute for professional fact-checking. Treat predictions as a signal, not a verdict.

Key Features

• Multi-Model Support: Choose between Logistic Regression, Naive Bayes, SVM, and DistilBERT.
• Explainability: Integrated feature importance and confidence scoring for every prediction.
• RESTful API: Built with FastAPI, providing high-performance endpoints for real-time inference.
• Extensible Pipeline: Modular data loaders and preprocessors for easy integration of new datasets.
• Interactive Frontend: A clean, responsive web interface for instant article analysis.
• Persistence: Built-in SQLite history tracking for auditing and analysis.

Architecture

TruthNet follows a modular architecture separating concerns between data processing, model management, and service delivery:

graph TD
    A[Raw Data / API Request] --> B[Preprocessing Engine]
    B --> C[Feature Extraction / Tokenization]
    C --> D{Model Factory}
    D --> E[Classical ML: Logistic/SVM/NB]
    D --> F[Deep Learning: DistilBERT]
    E --> G[Prediction Service]
    F --> G
    G --> H[FastAPI Endpoint]
    H --> I[Frontend UI / External Clients]
    G --> J[(SQLite History DB)]

Performance Benchmarks

The following metrics were achieved using the mrm8488/fake-news dataset (approx. 38,600 articles):

Model	Accuracy	Precision	Recall	F1-Score
DistilBERT	100.0%*	100.0%	100.0%	100.0%
SVM	99.34%	99.35%	99.20%	99.27%
Logistic Regression	98.67%	99.04%	98.01%	98.52%
Naive Bayes	95.33%	94.37%	95.34%	94.85%

*Note: DistilBERT was fine-tuned on a representative subset to optimize training efficiency.

Installation

Prerequisites

• Python 3.10+
• Virtual Environment (recommended)

Setup

# Clone the repository
git clone https://github.com/vincenzo-afk/TruthNet.git
cd TruthNet

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Configuration

TruthNet uses environment variables for configuration. Copy the example file and adjust as needed:

cp .env.example .env

Key Variables

Variable	Description	Default
MODEL_TYPE	Default model for training/API	logistic
API_PORT	Port for the FastAPI server	8000
DATASET_PATH	Path to the training CSV/JSON	datasets/dataset.csv

Training Pipeline

The training pipeline is fully automated and supports various model architectures.

# Train the default model (Logistic Regression)
python training/train.py

# Train a specific model with a custom dataset
python training/train.py --model svm --dataset path/to/data.csv

# Fine-tune the transformer model
python training/train.py --model distilbert

API Documentation

Once the server is running, interactive Swagger documentation is available at /docs.

Primary Endpoints

• GET /health: Check system and model status.
• POST /predict: Submit an article for classification.
• GET /history: Retrieve recent prediction history.

Sample Request

{
  "title": "Breaking News",
  "text": "Full article content goes here..."
}

Frontend Interface

TruthNet includes a built-in frontend accessible at the root URL (http://localhost:8000). It provides:

• A simple text area for article input.
• Real-time prediction visualization.
• Confidence meters and explainability highlights.

Contributing

We welcome contributions! Please follow these steps:

1. Fork the repository.
2. Create a feature branch (git checkout -b feature/AmazingFeature).
3. Commit your changes (git commit -m 'Add AmazingFeature').
4. Push to the branch (git push origin feature/AmazingFeature).
5. Open a Pull Request.

License

Distributed under the MIT License. See LICENSE for more information.

---

Developed by vincenzo-afk