Skip to content
Preprint

Not All Attacks Are Learned Equally in Speech Deepfake Detection

Sep 2026 · 0 citations · 46 references
Engineering

TL;DR

A replay-regularized, attack-aware curriculum that steps exposure based on measured attack influence is proposed that shows improved overall robustness and reduced attack-level imbalance compared with standard multi-attack training.

Abstract

Speech deepfake detection (SDD) models are trained on multi-attack datasets containing diverse spoofing systems, such as text-to-speech (TTS) and voice conversion (VC). In standard classifier training on multi-attack datasets, all attacks are treated as one spoofed class, and performance is reported using overall Equal Error Rate (EER). This aggregate view obscures how individual attacks shape learning and generalization. To better understand this attack-level behavior, we first balance TTS and VC exposure using sample and attack omission. We then measure attack-wise EER at inference and analyze attack-wise training loss and predictive entropy to characterize optimization. Results show that attacks contribute unequally: some attacks have high EER sensitivity and concentrated entropy with low loss, indicating strong influence on the decision boundary. We define these as high-impact attacks. To reduce uneven generalization across attacks, we propose a replay-regularized, attack-aware curriculum that steps exposure based on measured attack influence. Experiments on ASVspoof 2019, 2021, ASVspoof 5, and Fake-or-Real show improved overall robustness and reduced attack-level imbalance compared with standard multi-attack training.

View source

Similar papers

Open access Oct 2026

Defending against deepfake speech attacks via feature dropout and dropin regularization in large audio language models

Deepfakes have raised widespread concern owing to their threats to privacy, security, and societal trust, driving growing research interest in effective detection methods. Audio deepfake detection has moved from handcrafted-feature classifiers to end-to-end deep learning architectures and, more recently, to self-superv...

Yu-Pei Li, Yi-Xiong Fang, Jia-Hao Xue et al. · 0 citations
Open access 2026

Cross-Domain Generalization From Synthesized to Human-Imitated Speech Detection

The increasing prevalence of spoofing attacks in automatic speaker verification systems underscores the critical need for robust detection methods to ensure security and reliability. Although current spoof speech datasets primarily focus on synthesized speech and voice conversion attacks, human-imitated speech remains...

Khalid Zaman, Kai Li, Anuwat Chaiwongyen et al. · 0 citations
Preprint Aug 2026

Never Stop Speaking: a Denial-of-Service Attack on End-to-End Speech Language Models

The perturbation-based DoS attack targeting E2E speech models is proposed, formulated as a composite optimization objective that jointly suppresses EOS generation, encourages prolonged decoding, and largely preserves semantic consistency by integrating weighted EOS loss, top-k logit loss, length loss, and semantic alig...

Shuo Cheng, Kunlan Xiang, Ming-Xuan Li et al. · 0 citations
Review Open access Sep 2026

A survey of threats against voice authentication and anti-spoofing systems

Voice authentication has evolved from traditional systems based on hand-engineered acoustic features to deep learning models that learn speaker representations directly from audio. This progress has improved accuracy, but it has also opened new attack surfaces. Prior surveys treat these threats in isolation; we inste...

Kamel Kamel, Keshav Sood, Hridoy Sankar Dutta et al. · 4 citations
Open access 2026

Multi-Prototype Variational Information Bottleneck for SSL-Based Speech Spoofing Detection

Rapid advances in text-to-speech (TTS) and voice conversion (VC) have substantially lowered the barrier to generating zero-shot spoofed speech, posing critical challenges for identity authentication, media forensics, and speech security. Current audio deepfake detection methods typically employ large-scale self-supervi...

Heng-Chang Hou, Ruo-Hua Zhou, Qing-Sheng Yuan · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.