Deepfake Audio Detector

What It Does

Installation

git clone https://github.com/your-repo/deepfake_audio_detector.git
   cd deepfake_audio_detector

pip install -r requirements.txt

Usage

python deepfake_audio_detector.py --input sample_audio.mp3

Confidence Score: 75.32%
Likelihood: Deepfake

Source Code

import argparse
import os
from pydub import AudioSegment
import torch
import torchaudio

def analyze_audio(file_path):
    """Analyze the audio file for deepfake artifacts."""
    try:
        # Load the audio file
        audio = AudioSegment.from_file(file_path)
        samples = torch.tensor(audio.get_array_of_samples(), dtype=torch.float32)
        sample_rate = audio.frame_rate

        # Ensure the audio is mono
        if audio.channels > 1:
            raise ValueError("Audio file must be mono.")

        # Resample audio to 16kHz for analysis
        resampled_audio = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)(samples)

        # Placeholder for deepfake detection logic
        # Replace with actual AI model inference
        confidence_score = torch.rand(1).item() * 100
        likelihood = "Deepfake" if confidence_score > 50 else "Authentic"

        return confidence_score, likelihood

    except Exception as e:
        raise RuntimeError(f"Error analyzing audio file: {e}")

def main():
    parser = argparse.ArgumentParser(description="Deepfake Audio Detector")
    parser.add_argument("--input", required=True, help="Path to the audio file")
    args = parser.parse_args()

    input_path = args.input

    if not os.path.isfile(input_path):
        print(f"Error: File not found at {input_path}")
        return

    try:
        confidence_score, likelihood = analyze_audio(input_path)
        print(f"Confidence Score: {confidence_score:.2f}%")
        print(f"Likelihood: {likelihood}")
    except RuntimeError as e:
        print(e)

if __name__ == "__main__":
    main()