Deepfake Audio Detection Project

Overview

This project was developed during the AIAmplify Hackathon, a 24-hour hackathon focused on using AI to address real-world challenges. The goal of this project is to detect deepfake audio using machine learning techniques. The project uses MFCC (Mel-frequency cepstral coefficients) features extracted from audio files and a Support Vector Machine (SVM) classifier to differentiate between genuine and deepfake audio. Find the Paper here

Citation

A. Hamza et al., "Deepfake Audio Detection via MFCC Features Using Machine Learning," in IEEE Access, vol. 10, pp. 134018-134028, 2022, doi: 10.1109/ACCESS.2022.3231480. Abstract: Deepfake content is created or altered synthetically using artificial intelligence (AI) approaches to appear real. It can include synthesizing audio, video, images, and text. Deepfakes may now produce natural-looking content, making them harder to identify. Much progress has been achieved in identifying video deepfakes in recent years; nevertheless, most investigations in detecting audio deepfakes have employed the ASVSpoof or AVSpoof dataset and various machine learning, deep learning, and deep learning algorithms. This research uses machine and deep learning-based approaches to identify deepfake audio. Mel-frequency cepstral coefficients (MFCCs) technique is used to acquire the most useful information from the audio. We choose the Fake-or-Real dataset, which is the most recent benchmark dataset. The dataset was created with a text-to-speech model and is divided into four sub-datasets: for-rece, for-2-sec, for-norm and for-original. These datasets are classified into sub-datasets mentioned above according to audio length and bit rate. The experimental results show that the support vector machine (SVM) outperformed the other machine learning (ML) models in terms of accuracy on for-rece and for-2-sec datasets, while the gradient boosting model performed very well using for-norm dataset. The VGG-16 model produced highly encouraging results when applied to the for-original dataset. The VGG-16 model outperforms other state-of-the-art approaches. keywords: {Deepfakes;Deep learning;Speech synthesis;Training data;Feature extraction;Machine learning algorithms;Data models;Acoustics;Deepfakes;deepfake audio;synthetic audio;machine learning;acoustic data}, URL: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9996362&isnumber=9668973

Contributors to this project code

• Noor Chauhan
• Abhishek Khadgi
• Omkar Sapkal
• Himanshi Shinde
• Furqan Ali

Table of Contents

1. Introduction
2. Overview
3. Contributors
4. Installation
5. How to Use
   • Training the Model
   • Analyzing Audio
6. License

Installation

To initialize the project, follow these steps:

1. Clone the repository to your local machine:

   git clone https://github.com/your-username/deepfake-audio-detection.git
   cd deepfake-audio-detection
   

2. Set up a virtual environment (optional but recommended):

   # For Windows
   python -m venv venv
   venv\Scripts\activate
   
   # For Linux/macOS
   python3 -m venv venv
   source venv/bin/activate
   

3. Install the required dependencies

How to Use Training the Model

To train the SVM model with the provided data, follow these steps:

1. Prepare the dataset: Place genuine audio files in the real_audio directory and deepfake audio files in the deepfake_audio directory.

2. Run the training script:

   python main.py
   

   After sucessfully running the main script, it will initially ask you to provide the path of the voice to analyze, provide it with the path and the

3. Run the web app by:

   python app.py
   

   The training script will extract MFCC features from the audio files, split the data into training and testing sets, scale the features, train the SVM model, and save the trained model and scaler for future use.

Analyzing Audio

To classify an audio file as genuine or deepfake, follow these steps:

1. Ensure the trained model and scaler are available (already saved during training).

2. Run the analysis script:

   python analyze_audio.py path/to/your/audio/file.wav
   

   Replace path/to/your/audio/file.wav with the path to the audio file you want to analyze. The script will extract MFCC features from the audio, scale the features using the saved scaler, pass the features to the trained SVM model, and display the classification result.

Contribution & License

• For contributing, fork this project and compare and submit a pull request with proper description to your changed/added features
• OpenSource MIT License, for more information read the License.