@inproceedings{nguyen-etal-2026-analyzing,
title = "Analyzing Reasoning Shifts in Audio Deepfake Detection Under Adversarial Attacks",
author = "Nguyen, Binh and
Fleming, Charles and
Le, Thai",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.739/",
pages = "15512--15532",
ISBN = "979-8-89176-462-0",
abstract = "While Audio Language Models (ALMs) offer a promising shift towards explainable audio deepfake detection (ADD) via generated reasoning, whether these explanations reflect faithful perceptual grounding or fragile post-hoc rationalizations under adversarial attack remains under-explored. To address this gap, we introduce **SARA** (**S**hift **A**nalysis of **R**easoning in **A**udio), a diagnostic framework that evaluates the reasoning of five ALMs across three dimensions: acoustic perception, reasoning-verdict coherence and dissonance. Our experiments reveal three key findings. First, current ALMs exhibit weak acoustic perception ($< 50${\%} accuracy), and generating explicit rationales imposes an accuracy penalty (reducing real F1 score by up to 21{\%}). Second, adversarial attacks show a modality asymmetry: acoustic attacks severely degrade reasoning-verdict coherence (average drop of 14.20{\%}), whereas linguistic attacks achieve higher attack success (61.73{\%}) while preserving fluent coherence (a drop of only 3.91{\%}). Third, we demonstrate that decomposing reasoning shifts into prediction confidence and reasoning-verdict coherence enables a secondary defense layer that can detect adversarially perturbed audio with a 74.41{\%} F1 score even with unseen attack strategies and speaker profiles. These findings suggest that reasoning traces also provide diagnostic utility that persists even when final classification outputs are compromised. Our code is available at [https://github.com/nqbinh17/sara{\_}reasoning{\_}shifts](https://github.com/nqbinh17/sara{\_}reasoning{\_}shifts)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nguyen-etal-2026-analyzing">
<titleInfo>
<title>Analyzing Reasoning Shifts in Audio Deepfake Detection Under Adversarial Attacks</title>
</titleInfo>
<name type="personal">
<namePart type="given">Binh</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Charles</namePart>
<namePart type="family">Fleming</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thai</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>While Audio Language Models (ALMs) offer a promising shift towards explainable audio deepfake detection (ADD) via generated reasoning, whether these explanations reflect faithful perceptual grounding or fragile post-hoc rationalizations under adversarial attack remains under-explored. To address this gap, we introduce **SARA** (**S**hift **A**nalysis of **R**easoning in **A**udio), a diagnostic framework that evaluates the reasoning of five ALMs across three dimensions: acoustic perception, reasoning-verdict coherence and dissonance. Our experiments reveal three key findings. First, current ALMs exhibit weak acoustic perception (< 50% accuracy), and generating explicit rationales imposes an accuracy penalty (reducing real F1 score by up to 21%). Second, adversarial attacks show a modality asymmetry: acoustic attacks severely degrade reasoning-verdict coherence (average drop of 14.20%), whereas linguistic attacks achieve higher attack success (61.73%) while preserving fluent coherence (a drop of only 3.91%). Third, we demonstrate that decomposing reasoning shifts into prediction confidence and reasoning-verdict coherence enables a secondary defense layer that can detect adversarially perturbed audio with a 74.41% F1 score even with unseen attack strategies and speaker profiles. These findings suggest that reasoning traces also provide diagnostic utility that persists even when final classification outputs are compromised. Our code is available at [https://github.com/nqbinh17/sara_reasoning_shifts](https://github.com/nqbinh17/sara_reasoning_shifts).</abstract>
<identifier type="citekey">nguyen-etal-2026-analyzing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.739/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>15512</start>
<end>15532</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Analyzing Reasoning Shifts in Audio Deepfake Detection Under Adversarial Attacks
%A Nguyen, Binh
%A Fleming, Charles
%A Le, Thai
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F nguyen-etal-2026-analyzing
%X While Audio Language Models (ALMs) offer a promising shift towards explainable audio deepfake detection (ADD) via generated reasoning, whether these explanations reflect faithful perceptual grounding or fragile post-hoc rationalizations under adversarial attack remains under-explored. To address this gap, we introduce **SARA** (**S**hift **A**nalysis of **R**easoning in **A**udio), a diagnostic framework that evaluates the reasoning of five ALMs across three dimensions: acoustic perception, reasoning-verdict coherence and dissonance. Our experiments reveal three key findings. First, current ALMs exhibit weak acoustic perception (< 50% accuracy), and generating explicit rationales imposes an accuracy penalty (reducing real F1 score by up to 21%). Second, adversarial attacks show a modality asymmetry: acoustic attacks severely degrade reasoning-verdict coherence (average drop of 14.20%), whereas linguistic attacks achieve higher attack success (61.73%) while preserving fluent coherence (a drop of only 3.91%). Third, we demonstrate that decomposing reasoning shifts into prediction confidence and reasoning-verdict coherence enables a secondary defense layer that can detect adversarially perturbed audio with a 74.41% F1 score even with unseen attack strategies and speaker profiles. These findings suggest that reasoning traces also provide diagnostic utility that persists even when final classification outputs are compromised. Our code is available at [https://github.com/nqbinh17/sara_reasoning_shifts](https://github.com/nqbinh17/sara_reasoning_shifts).
%U https://aclanthology.org/2026.findings-emnlp.739/
%P 15512-15532
Markdown (Informal)
[Analyzing Reasoning Shifts in Audio Deepfake Detection Under Adversarial Attacks](https://aclanthology.org/2026.findings-emnlp.739/) (Nguyen et al., Findings 2026)
ACL