📌 Fake News Detection using RoBERTa-Based LLMs

A complete end-to-end fake news detection system with a fine-tuned RoBERTa model, backend API, and a Chrome Extension.

Demo

🚀 Project Overview

This project presents a RoBERTa-based fake news detection system that identifies misinformation from online article headlines. It includes:

✔ A fine-tuned RoBERTa model ✔ Backend Flask API for real-time predictions ✔ A Chrome Extension that users can install manually ✔ Scripts for downloading large model files from Google Drive

The system classifies news as FAKE or REAL with high accuracy and supports real-time browser-level predictions.

🗂 Dataset

Includes Fake.csv and True.csv containing:

Headlines Article text Subject Publish date

📊 Model Performance

RoBERTa gives the best accuracy for short and long news formats Provides better accuracy without high computation cost Balanced performance for real-time deployment

🧩 Chrome Extension

The Chrome Extension: ✔ Extracts news headline from any news page ✔ Sends it to your backend API ✔ Shows REAL / FAKE instantly ✔ Works in a popup window ✔ Requires no login or signup

How to Install the Extension

Go to chrome://extensions/ Turn ON Developer Mode Click Load Unpacked Select the extension/ folder from this repository Extension will appear in the Chrome toolbar

💾 Model Download (Google Drive)

GitHub cannot store files >100 MB, so download them automatically.

Python Script pip install gdown python download_models.py

Shell Script chmod +x download_models.sh ./download_models.sh

Downloads: best_roberta.pt roberta_model_weights.pt

Stored into: models/

Direct Links

best_roberta.pt https://drive.google.com/file/d/1pbdk6kV4tuAbAMKGZPfnGixH2cL6Ou18/view?usp=sharing

roberta_model_weights.pt https://drive.google.com/file/d/1TWH1x8RCXGtMTzcJtf-OBJWMA9IV3dsb/view?usp=sharing

🏗 Project Structure

FakeNewsDetector/ │── api/ │ ├── app.py │ ├── model/ │ ├── download_models.py │── extension/ │ ├── manifest.json │ ├── popup.html │ ├── popup.js │ ├── icon.png │── models/ (ignored) │── download_models.sh │── README.md │── requirements.txt

🚀 How to Run Backend API

Install dependencies pip install -r requirements.txt

Download models python download_models.py

Run Flask API python app.py

API runs on: http://localhost:5000/predict