Fake News Detector

Table of Contents

1. Overview
2. Features
3. Tech Stack
4. Project Structure
5. Model Training & Inference
6. Flask Backend Setup
7. Chrome Extension Setup
8. Deployment
9. Future Improvements
10. References

---

Overview

Fake News Detector is a project that identifies potentially fake or real news content using a Machine Learning (ML) model. It comes with a Chrome Extension that lets users highlight any text on a webpage and quickly get a classification result (FAKE vs. REAL), along with a confidence/distance score.

This project demonstrates:

1. Natural Language Processing (NLP) with TF-IDF.
2. Machine Learning model building (scikit-learn).
3. Flask for serving a RESTful prediction API.
4. Chrome Extension development (Manifest V3).
5. Deployment considerations (e.g., Heroku).

---

Features

• Train an ML model on a labeled dataset of fake and real news.
• Classify user-highlighted text on any webpage in real-time using a Chrome Extension.
• Confidence or Distance Score (depending on classifier capabilities).
• User-Friendly interface via the extension’s popup.

---

Tech Stack

• Python 3.7+
• Flask (for the backend)
• scikit-learn, pandas, numpy (for ML)
• Chrome Extension (Manifest V3)
• Heroku (optional deployment)

---

Project Structure

FakeNewsDetector/
├── fake_news_app/
│ ├── main.py # Flask app entry point
│ ├── model.py # Model training logic
│ ├── news.csv # Dataset
│ └── requirements.txt # Dependencies
└── chrome_extension/
├── manifest.json # Chrome extension descriptor (Manifest V3)
├── popup.html # Popup interface
├── popup.js # Chrome extension logic
├── background.js # (Optional) service worker or additional extension logic

In fake_news_app/

• model.py
  Contains the logic to train and evaluate the model using TfidfVectorizer and PassiveAggressiveClassifier.
• main.py
  A Flask app that loads or trains the model on startup, then exposes a /predict endpoint for inference.
• news.csv
  A sample dataset for fake/real news classification.
• requirements.txt
  Lists Python dependencies (Flask, scikit-learn, gunicorn, pandas, etc.).

In chrome_extension/

• manifest.json
  Chrome Extension configuration (permissions, name, version, etc.).
• popup.html
  The HTML UI displayed when clicking the extension’s icon.
• popup.js
  Handles user interactions (e.g. capturing highlighted text, calling Flask API).
• background.js
  An optional service worker file for background tasks.

---

Model Training & Inference

1. Training

   • We typically use model.py to train the model on news.csv.
   • Core steps:
     1. Load the dataset into a Pandas DataFrame.
     2. Split into training and test sets (e.g., 80/20).
     3. Vectorize the text with TfidfVectorizer.
     4. Train a PassiveAggressiveClassifier (or alternative).
     5. Evaluate accuracy and confusion matrix.

2. Inference

   • The Flask app (main.py) loads the trained vectorizer and model at startup.
   • /predict endpoint receives JSON input (text), vectorizes it, and returns a JSON response (FAKE or REAL + confidence/distance measure).

---

Flask Backend Setup

1. Install Requirements

   cd fake_news_app
   pip install -r requirements.txt
   

2. Train the Model (Optional if main.py automatically trains on startup, but you can also pre-train and save the model)

python model.py

Adjust your scripts accordingly if you’re saving/loading a model pickle.

3. Run the Flask App

python main.py

By default, the server starts at: http://127.0.0.1:5000. Test the Endpoint

Use curl or a tool like Postman: bash Copy code curl -X POST -H "Content-Type: application/json"
-d '{"text": "Suspicious news text"}'
http://127.0.0.1:5000/predict You should receive a JSON response with prediction and confidence. Chrome Extension Setup Load the Extension into Chrome

Go to chrome://extensions/. Enable Developer mode (toggle on top-right). Click Load unpacked. Select the chrome_extension/ folder. Configure the Endpoint

In popup.js, find the line where it calls the Flask endpoint: js Copy code const flaskUrl = "http://127.0.0.1:5000/predict"; If you deploy your Flask app (e.g. on Heroku), update this to your production URL. Usage

Navigate to any webpage. Highlight some text you want to check. Click on the Fake News Detector extension icon. Hit the “Check Selected Text” button. The popup will show whether the text is likely FAKE or REAL (and optionally a confidence score). Deployment Deploying to Heroku Make sure you have a Procfile in fake_news_app/: makefile Copy code web: gunicorn main:app Ensure your requirements.txt includes gunicorn. Create and push to Heroku: bash Copy code heroku create your-fake-news-app git init git add . git commit -m "Initial commit" heroku git:remote -a your-fake-news-app git push heroku master Once deployed, Heroku gives you a URL (e.g., https://your-fake-news-app.herokuapp.com). In your Chrome Extension (popup.js), update flaskUrl to: js Copy code const flaskUrl = "https://your-fake-news-app.herokuapp.com/predict"; Future Improvements Probability Scores

Consider using LogisticRegression or SGDClassifier(loss='log') to get probability outputs (predict_proba). Advanced NLP

Integrate data-cleaning, lemmatization, or advanced embeddings (e.g., Word2Vec, BERT). Frontend Enhancement

Provide a more polished UI in the Chrome Extension, or highlight the text in-page with color-coded results. Caching & Scalability

For higher traffic, consider caching results or scaling the Flask app with a production-grade server and database. References Detecting Fake News with Python and Machine Learning - DataFlair How to Build a Fake News Classification Model (ODS) Chrome Extension Official Docs Flask Documentation Heroku Deployment Guide Author: [Your Name or Team Name] LinkedIn | Portfolio | Twitter

Feel free to star this repository if you found it helpful! If you have any questions, please open an issue or reach out.

License: Choose an appropriate license, e.g. MIT License.

Thank you for checking out the Fake News Detector!