Introduction to Voice Presentation Attack Detection and Recent Advances

Sahidullah, Md; Delgado, Héctor; Todisco, Massimiliano; Kinnunen, Tomi; Evans, Nicholas; Yamagishi, Junichi; Lee, Kong Aik

doi:10.1007/978-981-19-5288-3_13

Cited by 5 publications

(2 citation statements)

References 190 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite the advantages synthetic speech offers, the rise of audio deepfakes presents a severe threat to our society and economy. Deepfake audio employs advanced deep learning techniques like Text-to-Speech (TTS) or Voice Conversion (VC) to generate sophisticated spoofed audio samples, which can be considered presentation attacks [1]. Such maliciously crafted artificial speech can deceive both Automatic speaker verification systems (ASV), posing a severe threat to the integrity and reliability of speaker verification processes.…”

Section: Introductionmentioning

confidence: 99%

“…Due to the transferability nature across deep learning-based models [18], [19], adversarial samples are often retrieved from a surrogate (substitute) model to attack the target (supposedly similar) model. As announced by the ASVspoof organizers, ASVspoof5 competition 1 will shift the focus to VC and TTS, including adversarial attacks relying on ASV/CM feedback. The need to build resilient defense mechanisms has never been more urgent, and addressing these challenges requires our utmost attention.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

On the Defense of Spoofing Countermeasures Against Adversarial Attacks

Nguyen-Vu,

Doan,

Bui

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Advances in speech synthesis have exposed the vulnerability of spoofing countermeasure (CM) systems. Adversarial attacks exacerbate this problem, mainly due to the reliance of most CM models on deep neural networks. While research on adversarial attacks in anti-spoofing systems has received considerable attention, there is a relative scarcity of studies focused on developing effective defense techniques. In this study, we propose a defense strategy against such attacks by augmenting training data with frequency bandpass filtering and denoising. Our approach aims to limit the impact of perturbation, thereby reducing the susceptibility to adversarial samples. Furthermore, our findings reveal that the use of Max-Feature-Map (MFM) and frequency band-pass filtering provides additional benefits in suppressing different noise types. To empirically validate this hypothesis, we conduct tests on different CM models using adversarial samples derived from the ASVspoof challenge and other well-known datasets. The evaluation results show that such defense mechanisms can potentially enhance the performance of spoofing countermeasure systems.

show abstract

Section: Introductionmentioning

confidence: 99%