RW-Resnet: A Novel Speech Anti-Spoofing Model Using Raw Waveform

Ma, Youxuan; Ren, Zongze; Xu, Shugong

doi:10.48550/arxiv.2108.05684

Cited by 4 publications

(4 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The RW-Resnet is method presented by Ma et al [218] for speech anti-spoofing that works directly on raw waveforms. This approach contrasts with traditional techniques, which often rely on spectrograms or other handcrafted features.…”

Section: ) Methods Using Deep Learningmentioning

confidence: 99%

“…On the other hand, deep learning feature-based methods, such as Attentive Filtering Networks (AFN) by Lai et al [215], Deep-Feature Extractor for Automatic Speaker Verification (ASV) spoofing detection by Gomez-Alanis et al [216], and raw-waveform processing strategy by Ma et al [218], have demonstrated high detection rates for audio deepfakes. These methods effectively leverage intricate non-linear relationships and temporal dependencies in audio signals, therby maximising the capabilities of deep learning algorithms.…”

Section: Audio Deepfake Detectionmentioning

confidence: 99%

See 1 more Smart Citation

A Survey on the Detection and Impacts of Deepfakes in Visual, Audio, and Textual Formats

Mubarak,

Alsboui,

Alshaikh

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In the rapidly evolving digital landscape, the generation of fake visual, audio, and textual content poses a significant threat to society's trust, political stability, and integrity of information. The generation process has been enhanced and simplified using Artificial Intelligence techniques, which have been termed deepfake. Although significant attention has been paid to visual and audio deepfakes, there is also a burgeoning need to consider text-based deepfakes. Due to advancements in natural language processing and large language models, the potential of manipulating textual content to reshape online discourse and misinformation has increased. This study comprehensively examines the multifaceted nature and impacts of deep-fake-generated media. This work explains the broad implications of deepfakes in social, political, economic, and technological domains. State-of-the-art detection methodologies for all types of deepfake are critically reviewed, highlighting the need for unified, real-time, adaptable, and generalised solutions. As the challenges posed by deepfakes intensify, this study underscores the importance of a holistic approach that intertwines technical solutions with public awareness and legislative action. By providing a comprehensive overview and establishing a framework for future exploration, this study seeks to assist researchers, policymakers, and practitioners in navigating the complexities of deepfake phenomena.

show abstract

Section: ) Methods Using Deep Learningmentioning

confidence: 99%

Section: Audio Deepfake Detectionmentioning

confidence: 99%

A Survey on the Detection and Impacts of Deepfakes in Visual, Audio, and Textual Formats

Mubarak,

Alsboui,

Alshaikh

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…However, this method has not been tested across corpora. Following this, the author of [38] proposes a Conv1D Resblock with a residual connection, which allows the model to learn a better feature representation from raw waveforms. They find that feeding a raw waveform directly into a neural network is adequate.…”

Section: Deep Learning Solutionsmentioning

confidence: 99%

Battling voice spoofing: a review, comparative analysis, and generalizability evaluation of state-of-the-art voice spoofing counter measures

et al. 2023

View full text Add to dashboard Cite

With the advent of automated speaker verification (ASV) systems comes an equal and opposite development: malicious actors may seek to use voice spoofing attacks to fool those same systems. Various countermeasures have been proposed to detect these spoofing attacks, but current offerings in this arena fall short of a unified solution. For this reason, defensive measures for ASV systems produced in the last 6-7 years need to be classified, and qualitative and quantitative comparisons of state-of-the-art (SOTA) countermeasures should be performed to assess the effectiveness of these systems against real-world attacks. Hence, in this work, we conduct a review of the literature on spoofing detection using hand-crafted features, deep learning, and end-to-end spoofing countermeasure solutions to detect logical access (LA) attacks, such as speech synthesis (SS) and voice conversion (VC), and physical access (PA) attacks, i.e., replay attacks. Additionally, we review integrated and unified solutions to voice spoofing evaluation and speaker verification, and adversarial and anti-forensic attacks on both voice countermeasures and ASV systems. In an extensive experimental analysis, the limitations and challenges of existing spoofing countermeasures are presented, the performance of these countermeasures on several datasets is reported, and cross-corpus evaluations are performed, something that is nearly absent in the existing literature, in order to assess the generalizability of existing solutions. For the experiments, we employ the Voice Spoofing Attacks and Countermeasures ASVspoof2019, ASVspoof2021, and VSDC datasets along with GMM, SVM, CNN, and CNN-GRU classifiers. (For reproducibility of the results, the code of the testbed can be found at our GitHub Repository * ).

show abstract

“…Recent studies proposed end-to-end anti-spoofing systems that operate on raw waveforms [17,18,19]. These end-toend models can also be viewed as a category of methods that replace the aforementioned feature extraction with neural nets with trainable parameters.…”

Section: Introductionmentioning

confidence: 99%

Representation Selective Self-distillation and wav2vec 2.0 Feature Exploration for Spoof-aware Speaker Verification

Lee,

Kim,

Koo

et al. 2022

Preprint

View full text Add to dashboard Cite

Text-to-speech and voice conversion studies are constantly improving to the extent where they can produce synthetic speech almost indistinguishable from bona fide human speech. In this regrad, the importance of countermeasures (CM) against synthetic voice attacks of the automatic speaker verification (ASV) systems emerges. Nonetheless, most end-to-end spoofing detection networks are black box systems, and the answer to what is an effective representation for finding artifacts still remains veiled. In this paper, we examine which feature space can effectively represent synthetic artifacts using wav2vec 2.0, and study which architecture can effectively utilize the space. Our study allows us to analyze which attribute of speech signals is advantageous for the CM systems. The proposed CM system achieved 0.31% equal error rate (EER) on ASVspoof 2019 LA evaluation set for the spoof detection task. We further propose a simple yet effective spoofing aware speaker verification (SASV) methodology, which takes advantage of the disentangled representations from our countermeasure system. Evaluation performed with the SASV Challenge 2022 database show 1.08% of SASV EER. Quantitative analysis shows that using the explored feature space of wav2vec 2.0 advantages both spoofing CM and SASV.

show abstract

RW-Resnet: A Novel Speech Anti-Spoofing Model Using Raw Waveform

Cited by 4 publications

References 23 publications

A Survey on the Detection and Impacts of Deepfakes in Visual, Audio, and Textual Formats

A Survey on the Detection and Impacts of Deepfakes in Visual, Audio, and Textual Formats

Battling voice spoofing: a review, comparative analysis, and generalizability evaluation of state-of-the-art voice spoofing counter measures

Representation Selective Self-distillation and wav2vec 2.0 Feature Exploration for Spoof-aware Speaker Verification

Contact Info

Product

Resources

About