AUDI-FAKE: An Exploration into Intelligent Audio Deepfake Detection Foundation Models

This book presents AUDI-FAKE as a unified approach to intelligent audio deepfake detection using pretrained and foundation-model-based representations. It shows why no single pretrained model captures all the cues needed for reliable detection in diverse, realistic conditions and demonstrates how speech encoders, multilingual models, and audio-tagging models can be combined to improve performance. The book offers a progressive treatment of the subject, covering the foundations of audio deepfake detection, benchmark datasets, pretrained models, and conventional detection methods, while examining important and underexplored detection settings. It is a useful resource for researchers and advanced students working in speech, audio, and multimodal machine learning.