Audio DeepFake Detection using CNN-BiLSTM

APP Demo

Audio-DeepFake-Demo.mp4

Overview

This project aims to detect audio deepfakes using a hybrid approach that combines Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory Networks (BiLSTM). The system is designed to effectively classify audio data into genuine or fake categories, offering a robust solution to the growing challenges posed by audio-based misinformation.

---

Key Features

• Hybrid Model Architecture: Combines the feature extraction power of CNNs with the sequential processing capabilities of BiLSTMs.
• State-of-the-Art Accuracy: Achieves high detection accuracy, making it suitable for practical applications.
• Research Contribution: Includes detailed insights and a research paper explaining the methodology and findings.

---

Table of Contents

• Project Overview
• Key Features
• Dataset
• Model Architecture
• Installation
• Results
• Future Work
• Contributing

---

Dataset

• For Real Audio : https://www.kaggle.com/datasets/mathurinache/the-lj-speech-dataset
• For Fake Audio : https://www.kaggle.com/datasets/andreadiubaldo/wavefake-test
  • The dataset includes audio recordings, labeled as either genuine or deepfake.
  • Preprocessing steps involve:
    • Feature extraction using Mel-frequency cepstral coefficients (MFCCs).
    • Data augmentation techniques to enhance model robustness.

---

Model Architecture

The model leverages the strengths of:

1. CNN:
   • Extracts spatial features from MFCCs.
   • Efficiently identifies patterns and anomalies.
2. BiLSTM:
   • Processes sequential data to capture temporal dependencies.
   • Bidirectional design ensures both past and future context is utilized.

---

Installation

1. Clone the repository:

   git clone https://github.com/VivekShinde7/Audio-DeepFake-Detection-using-CNN-BiLSTM.git
   cd Audio-DeepFake-Detection-using-CNN-BiLSTM

2. Create a virtual environment (optional but recommended):

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

3. Install dependencies:

   pip install -r requirements.txt

4. Run app.py:

   streamlit run app.py

---

Results

• Performance Metrics:
  • Accuracy: 98.3%
  • Precision: 97.8%
  • Recall: 98.8%
• Visualization of confusion matrix, System Architecture & Evaluation is available in the results folder.

---

Future Work

• Enhance the dataset to include diverse languages and accents.
• Optimize the model for real-time detection.
• Explore the integration of transformer-based architectures like Wav2Vec2.0.

---

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository.
2. Create a feature branch:

   git checkout -b feature-name

3. Commit your changes:

   git commit -m "Add your message here"

4. Push to the branch:

   git push origin feature-name

5. Create a pull request.

---

Acknowledgments

• Special thanks to open-source contributors and dataset providers.
• Inspiration drawn from advancements in audio deepfake detection research.

For queries or suggestions, feel free to open an issue or contact Vivek Shinde.