Trustify: Fuzzy Logic-Based Hybrid Framework for Detecting Fake News

This repository is a complete implementation of the research paper "A novel fuzzy logic-based hybrid framework for detecting fake news"

Overview

Trustify is a multimodal fake news detection system that combines:

• Text Processing: LSTM networks for sequential text analysis
• Image Processing: ResNet50 + GNN + Transformer for visual feature extraction
• Fuzzy Logic Decision Layer: Interpretable reasoning for final classification
• Text-Image Similarity Module: Semantic alignment verification

Key Features

✅ End-to-end pipeline: crawling → preprocessing → training → evaluation
✅ Multimodal fusion with explainability
✅ Supports 3 benchmark datasets: Twitter, BuzzFeed, PolitiFact
✅ Achieves >96% accuracy on Twitter dataset
✅ Ready for Colab/GPU training
✅ Publication-quality visualizations

Performance Metrics (Paper Results)

Dataset	Accuracy	Precision	Recall	F1-Score
Twitter	96.1%	98.4%	97.2%	97.8%
BuzzFeed	95.9%	94.9%	92.2%	93.5%
PolitiFact	92.2%	94.9%	92.3%	93.9%

Quick Start

Prerequisites

• Python 3.9+
• CUDA 12.1 (for GPU, optional but recommended)
• 8GB RAM minimum (16GB+ for comfortable training)

Installation

# Clone repository
git clone https://github.com/fatimasood/NeuroVerify.git

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

# Install dependencies
pip install -r requirements.txt

# Download NLTK data
python -c "import nltk; nltk.download('punkt'); nltk.download('averaged_perceptron_tagger')"

1. Download/Prepare Datasets

The paper uses publicly available datasets. Download from:

# Create data directories
mkdir -p data/{twitter,buzzfeed,politifact}/{raw,processed}

# Twitter dataset
# Download from: https://www.kaggle.com/datasets/sudishsharma/twitter-fake-news-dataset

# BuzzFeed dataset
# Download from: https://www.kaggle.com/datasets/rmisra/buzzfeed-articles-fake-real-news

# PolitiFact dataset
# Download from: https://www.kaggle.com/datasets/rushi883/politifact-fake-real-dataset

2. Preprocess Data

python src/preprocessing/preprocess.py \
    --dataset twitter \
    --input_path data/twitter/raw \
    --output_path data/twitter/processed \
    --image_size 224

3. Train Model

python src/training/trainer.py \
    --dataset twitter \
    --epochs 20 \
    --batch_size 20 \
    --learning_rate 0.005 \
    --device cuda:0 \
    --output_dir checkpoints/

4. Evaluate Model

python src/evaluation/evaluate.py \
    --checkpoint checkpoints/best_model.pth \
    --dataset twitter \
    --batch_size 32 \
    --output_dir results/

5. Generate Visualizations

python src/utils/visualization.py \
    --results_dir results/ \
    --output_dir plots/

Training on Google Colab

Open notebooks/colab_train.ipynb and:

1. Click "Open in Colab"
2. Run cells sequentially
3. GPU will be automatically enabled
4. Results and plots saved to Colab environment

# In Colab:
!git clone https://github.com/fatimasood/NeuroVerify.git
!pip install -r requirements.txt
# Then follow notebook cells