Analyzing Reasoning Shifts in Audio Deepfake Detection under Adversarial Attacks: The Reasoning Tax versus Shield Bifurcation

2026-01-07

arXiv.org

AI Summary

This work addresses a critical gap in current audio deepfake detection methods, which predominantly emphasize the stability of prediction outcomes while overlooking the robustness of model reasoning under adversarial attacks. The authors propose a forensic auditing framework that systematically evaluates the reasoning robustness of audio-language models across three dimensions: acoustic perception, cognitive consistency, and cognitive dissonance. Their analysis reveals, for the first time, that reasoning plays a dual role in deepfake detection—acting both as a “shield” that enhances defense and as a “tax” that introduces vulnerabilities. Notably, elevated cognitive dissonance is identified as a silent alarm signaling potential manipulation. Experiments demonstrate that strong acoustic perception can leverage reasoning to improve detection, whereas language-level adversarial attacks degrade cognitive consistency and increase attack success; crucially, high cognitive dissonance remains an effective warning signal even when classification fails.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity

Abstract

Audio Language Models (ALMs) offer a promising shift towards explainable audio deepfake detections (ADDs), moving beyond \textit{black-box} classifiers by providing some level of transparency into their predictions via reasoning traces. This necessitates a new class of model robustness analysis: robustness of the predictive reasoning under adversarial attacks, which goes beyond existing paradigm that mainly focuses on the shifts of the final predictions (e.g., fake v.s. real). To analyze such reasoning shifts, we introduce a forensic auditing framework to evaluate the robustness of ALMs'reasoning under adversarial attacks in three inter-connected dimensions: acoustic perception, cognitive coherence, and cognitive dissonance. Our systematic analysis reveals that explicit reasoning does not universally enhance robustness. Instead, we observe a bifurcation: for models exhibiting robust acoustic perception, reasoning acts as a defensive \textit{``shield''}, protecting them from adversarial attacks. However, for others, it imposes a performance \textit{``tax''}, particularly under linguistic attacks which reduce cognitive coherence and increase attack success rate. Crucially, even when classification fails, high cognitive dissonance can serve as a \textit{silent alarm}, flagging potential manipulation. Overall, this work provides a critical evaluation of the role of reasoning in forensic audio deepfake analysis and its vulnerabilities.

Problem

• audio deepfake detection
• adversarial attacks
• reasoning robustness
• audio language models
• cognitive coherence

Innovation

• reasoning robustness
• audio deepfake detection
• adversarial attacks
• cognitive dissonance
• explainable AI