Classical Machine Learning Baselines for Deepfake Audio Detection on the Fake-or-Real Dataset

TL;DR

This paper develops and evaluates simple, interpretable classical machine learning models using acoustic features to detect deepfake audio, providing a transparent baseline for future research.

Contribution

It introduces a classical machine learning baseline with detailed feature analysis and statistical validation for deepfake audio detection on the Fake-or-Real dataset.

Findings

RBF SVM achieves ~93% accuracy and 7% EER.

Spectral features like spectral centroid and bandwidth are key discriminative cues.

Linear models reach around 75% accuracy.

Abstract

Deep learning has enabled highly realistic synthetic speech, raising concerns about fraud, impersonation, and disinformation. Despite rapid progress in neural detectors, transparent baselines are needed to reveal which acoustic cues reliably separate real from synthetic speech. This paper presents an interpretable classical machine learning baseline for deepfake audio detection using the Fake-or-Real (FoR) dataset. We extract prosodic, voice-quality, and spectral features from two-second clips at 44.1 kHz (high-fidelity) and 16 kHz (telephone-quality) sampling rates. Statistical analysis (ANOVA, correlation heatmaps) identifies features that differ significantly between real and fake speech. We then train multiple classifiers -- Logistic Regression, LDA, QDA, Gaussian Naive Bayes, SVMs, and GMMs -- and evaluate performance using accuracy, ROC-AUC, EER, and DET curves. Pairwise McNemar's tests confirm statistically significant differences between models. The best model, an RBF SVM, achieves ~93% test accuracy and ~7% EER on both sampling rates, while linear models reach ~75% accuracy. Feature analysis reveals that pitch variability and spectral richness (spectral centroid, bandwidth) are key discriminative cues. These results provide a strong, interpretable baseline for future deepfake audio detectors.