An Initial Study of Physiological Markers for Deepfake Speech Detection

Vojtěch Staněk; Michal Frič; Kamil Malinka; Anton Firc

cs.SDOct 13, 2026

An Initial Study of Physiological Markers for Deepfake Speech Detection

Authors: Vojtěch Staněk, Michal Frič, Kamil Malinka, Anton Firc

Abstract

Conventional deepfake speech detectors primarily target synthesis artifacts and often degrade under unseen attacks or recording conditions. Human speech, however, contains physiological markers associated with respiration, phonation, and temporal coordination arising from the speech-production system that synthesizers may not reproduce faithfully. This motivates the hypothesis that deepfake speech can be detected by identifying inconsistencies in these patterns, which we test using 36 handcrafted physiological markers organized into respiratory, vocal, and temporal categories. We evaluate the marker vectors with logistic regression and SVM classifiers on pooled English recordings. The best in-domain system achieves a 21.40% equal error rate (EER), while logistic regression achieves 32.77% on the out-of-domain dataset pair. Since the markers describe specific speech-production properties, the logistic-regression coefficients indicate whether higher marker values push the classifier toward bona fide or deepfake speech. These results provide initial evidence that physiological markers offer a useful, interpretable signal for deepfake speech detection.

Figures & tables

Category	Marker	Measures
Respiratory (14 markers)	Breath frequency	frequency
	Breath duration	mean, std, min, max
	Breath acoustics	spectral centroid, spectral bandwidth, spectral peak frequency
	Pre-speech inhalation	pre-speech inhalation ratio
	Speech-breath coordination	coordination score, speech-to-breath mean/std, breath-to-speech mean/std
Vocal (16 markers)	Fundamental frequency ( $F_{0}$ )	mean, std, min, max

TABLE I: Physiologically motivated categorization of the handcrafted markers in this work.

Panel of Figure 2 — Fig. 1: Overview of the proposed physiological-marker detector (best viewed in color). The input recording is represented using respiratory, vocal, and temporal markers. The resulting 36-dimensional vector is processed by a linear classifier, whose coefficients provide marker-level interpretability.

Dataset	Bona fide	Deepfake
ASVspoof 5 [ wang24_asvspoof ]	9 208	20 121
Codecfake [ codecfake ]	0	79
In-the-Wild [ muller2022inthewild ]	194	421
M-AILABS [ MAILABS_2017 ] + MLAAD [ muller24_mlaad ]	849	1 159
SCDF [ scdf ]	0	1 373
SpeechFake [ speechfake ]	1 947	46 932

TABLE II: Composition of the aggregated EDS-14 dataset (English recordings longer than 14 s).

Setting	Train	Eval.	LR	RBF SVM
ID	EDS-14 all train	EDS-14 all test	27.60	21.40
ID	EDS-14 ∖MLAAD train	EDS-14 ∖MLAAD test	29.47	41.25
OOD	EDS-14 ∖MLAAD train	MLAAD/ M-AILABS	32.77	37.83

TABLE III: Standalone physiological-marker detector EER [%]. EDS-14 all denotes full EDS-14 and EDS-14 ∖MLAAD denotes EDS-14 without MLAAD/M-AILABS.

Marker	Category	Coeff.	Supports
Shimmer (APQ5)	Vocal	$-4.01$	Bona fide
Mean breath duration	Respiratory	$+3.73$	Deepfake
Shimmer (absolute)	Vocal	$+3.29$	Deepfake
Articulation rate	Temporal	$+2.51$	Deepfake
Jitter (PPQ5)	Vocal	$+1.40$	Deepfake
Shimmer (APQ11)	Vocal	$+0.91$	Deepfake

TABLE IV: Most influential markers of the ID LR model, ranked by coefficient magnitude. Red/green indicates that higher-than-average values support the deepfake/bona fide class, respectively.

Explore similar work

Oct 13, 2026

Physics-Aware Deepfake Detection via Distance-Speech Consistency

Recent progress in generative models has made talking-head deepfakes highly realistic and easy to create, jeopardizing public trust. However, most audio-visual deepfake detectors rely on lip-speech synchronization and are designed for static, frontal videos, limiting their reliability in dynamic, in-the-wild recordings where the speaker moves and lip cues are degraded or missing. As an alternative, we propose a physics-aware detector for dynamic speaking videos that leverages an acoustic constraint: in real recordings, speech energy varies predictably with speaker-to-camera distance, whereas deepfake generation can break this coupling. Our method combines distance estimates from video with distance-dependent acoustic measures from speech to detect deepfakes. Experiments on in-the-wild benchmarks show that the proposed physical cue is effective in dynamic scenarios, and that it complements lip-sync-based detectors, with a simple ensemble achieving consistent gains across datasets. Codes are available at https://byulharang.github.io/PADD/.

Kyeongrae Kim, Kim Sung-Bin, Oh Hyun-Bin +1

School of Electrical Engineering, KAIST, Korea · Dept. of Electrical Engineering, POSTECH, Korea · School of Computing, KAIST, Korea

Deepfake Detection Audio Deepfake Detection

Ask PDF

Sep 8, 2026

From Scores to Evidence: Auditable Decisions Can Improve Speech Deepfake Detection

Speech deepfake detectors usually emit one score per utterance, but a borderline score does not reveal why two examples differ during retrospective error analysis. We ask whether a final score can be calibrated from component fields while keeping those fields visible for inspection. We build a decision record with a passive detector score and a score from a probe applied to a marked copy. It also includes retrieval support held out of the evaluated family, a margin from a support-set profile, and raw neighbor closeness. A cross-fit calibrator combines these fields and two differences between raw scores into one final score. On matched ASVspoof development data, the calibrated record reduces equal error rate (EER) by 3.48 percentage points relative to the fixed retrieval-augmented rule. It reaches 8.43% EER, whereas a passive WavLM baseline reaches 6.71% on the same subset. The record is therefore not the strongest detector in this comparison. Its value is to retain inspectable component fields while producing one scalar score for retrospective diagnosis.

Mengzhe Geng, Yujia Lu, Manuela Kunz +1

National Research Council Canada · The Chinese University of Hong Kong

Audio Deepfake Detection Explainable Deepfake Detection ASVspoof

Ask PDF

Aug 20, 2026

Tracking the Trend in How Speech Synthesizers Deceive People

Advances in speech synthesis have made deepfake audio highly realistic. Earlier studies reported 70-80% human detection accuracy, but relied primarily on older synthesizers. We compare human detection for three selected voice synthesis tools released in 2019, 2022, and 2024 with 82 IT professionals, and benchmark humans against six pretrained detectors on the same material. For fully synthetic speech (full spoofs), the F1 score drops from about 90% for RTVC and YourTTS to 48% for ElevenLabs, although listeners were explicitly warned that deepfakes were present. For partial spoofing, where only one sentence of an utterance is altered, strict accuracy falls to 9%, and listeners classify the synthetic sentence as bona fide 77% of the time. Humans and detectors fail in complementary ways, and neither reliably localizes short manipulations. Additionally, listeners increasingly mislabel bona fide speech as fake, eroding trust in unmanipulated audio. These findings show that human perception alone is unreliable for the selected modern and partial-spoof conditions and motivate procedural verification, provenance, watermarking, and segment-level detection.

Milan Šalko, Anton Firc, Kamil Malinka +4

Audio Deepfake Detection TTS Synthesis

Ask PDF