IAudioDeepfakeDetector<T>

Interfaces · AiDotNet.Safety.Audio

Interface for audio deepfake and voice cloning detection modules.

Implements: IAudioSafetyModule<T>

Known implementations: AudioDeepfakeDetectorBase<T>

For Beginners

An audio deepfake detector checks if a voice recording is real or AI-generated. It can detect cloned voices, synthesized speech, and voice conversion attacks by analyzing subtle patterns in the audio signal.

How It Works

Audio deepfake detectors analyze audio waveforms for signs of AI-generated speech, voice cloning, or voice conversion. Approaches include spectral analysis of mel spectrograms, speaker embedding verification, and watermark detection.

References:

  • SafeEar: Privacy-preserving audio deepfake detection (ACM CCS 2024)
  • VoiceRadar: Voice deepfake detection framework (NDSS 2025)
  • AudioSeal: Localized watermarking for voice cloning detection (Meta AI, 2024)

Methods

Method Summary
double GetDeepfakeScore(Vector<T> audioSamples, int sampleRate) Gets the deepfake probability score for the given audio (0.0 = authentic, 1.0 = fake).