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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ahmad, Faheem, Ahmed, Ajan, Imtiaz, Masudul
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911594540171264
author Ahmad, Faheem
Ahmed, Ajan
Imtiaz, Masudul
author_facet Ahmad, Faheem
Ahmed, Ajan
Imtiaz, Masudul
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2604_13400
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Classical Machine Learning Baselines for Deepfake Audio Detection on the Fake-or-Real Dataset
Ahmad, Faheem
Ahmed, Ajan
Imtiaz, Masudul
Audio and Speech Processing
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.
title Classical Machine Learning Baselines for Deepfake Audio Detection on the Fake-or-Real Dataset
topic Audio and Speech Processing
url https://arxiv.org/abs/2604.13400