Voice Conversion Attacks Research Articles

A Self-Attentional ResNet-LightGBM Model for IoT-Enabled Voice Liveness Detection

Voice user interface (VUI) brings high efficiency and convenience for the applications of Internet of Things (IoTs), meanwhile, it can also cause increasingly serious security issues. The word-level voice liveness detection is proved to be the promising solution to thwart spoofing attacks. However, The complex acoustic feature, diversified attacks, and different interaction distance can severely affects the improvement of detection accuracy. To alleviate this issue, we develop a novel pop noise based word-level voice liveness detection framework. Firstly, a new voice frame selection method is proposed for determining optimal frames, including short time Fourier transform, low frequency average energy computation and sequencing. Then, the acoustic features of the selected frames are calculated by Gammatone frequency cepstral coefficient (GFCC). Finally, based on these features, a newly-built joint voice detector, fusing the self-attentional residual network (ResNet) and light gradient boosting machine (LightGBM), can achieve accurate voice classification. On the popular voice spoofing attack datasets, experimental results show that our proposal significantly outperforming the baseline and the state of the arts models, and it is gender dependent. Moreover, our proposal has good generalization ability for far-field replay voice attack, speech synthesis and voice conversion attacks, and partial fake voice attack. Finally, its effectiveness is verified by the ablation study.

Artificial speech detection using image-based features and random forest classifier

The ASVspoof 2015 Challenge was one of the efforts of the research community in the field of speech processing to foster the development of generalized countermeasures against spoofing attacks. However, most countermeasures submitted to the ASVspoof 2015 Challenge failed to detect the S10 attack effectively, the only attack that was generated using the waveform concatenation approach. Hence, more informative features are needed to detect previously unseen spoofing attacks. This paper presents an approach that uses data transformation techniques to engineer image-based features together with random forest classifier to detect artificial speech. The objectives are two-fold: (i) to extract image-based features from the melfrequency cepstral coefficients representation of the speech signal and (ii) to compare the performance of using the extracted features and Random Forest to determine the authenticity of voices with the existing approaches. An audio-to-image transformation technique was used to engineer new features in classifying genuine and spoof voices. An experiment was conducted to find the appropriate combination of the engineered features and classifier. Experimental results showed that the proposed approach was able to detect speech synthesis and voice conversion attacks effectively, with an equal error rate of 0.10% and accuracy of 99.93%.

ASVspoof 2019: Spoofing Countermeasures for the Detection of Synthesized, Converted and Replayed Speech

The ASVspoof initiative was conceived to spearhead research in anti-spoofing for automatic speaker verification (ASV). This paper describes the third in a series of bi-annual challenges: ASVspoof 2019. With the challenge database and protocols being described elsewhere, the focus of this paper is on results and the top performing single and ensemble system submissions from 62 teams, all of which out-perform the two baseline systems, often by a substantial margin. Deeper analyses shows that performance is dominated by specific conditions involving either specific spoofing attacks or specific acoustic environments. While fusion is shown to be particularly effective for the logical access scenario involving speech synthesis and voice conversion attacks, participants largely struggled to apply fusion successfully for the physical access scenario involving simulated replay attacks. This is likely the result of a lack of system complementarity, while oracle fusion experiments show clear potential to improve performance. Furthermore, while results for simulated data are promising, experiments with real replay data show a substantial gap, most likely due to the presence of additive noise in the latter. This finding, among others, leads to a number of ideas for further research and directions for future editions of the ASVspoof challenge.

A Kernel Density Estimation Based Loss Function and its Application to ASV-Spoofing Detection

Biometric systems are exposed to spoofing attacks which may compromise their security, and voice biometrics, also known as automatic speaker verification (ASV), is no exception. Replay, synthesis and voice conversion attacks cause false acceptances that can be detected by anti-spoofing systems. Recently, deep neural networks (DNNs) which extract embedding vectors have shown superior performance than conventional systems in both ASV and anti-spoofing tasks. In this work, we develop a new concept of loss function for training DNNs which is based on kernel density estimation (KDE) techniques. The proposed loss functions estimate the probability density function (pdf) of every training class in each mini-batch, and compute a log likelihood matrix between the embedding vectors and pdfs of all training classes within the mini-batch in order to obtain the KDE-based loss. To evaluate our proposal for spoofing detection, experiments were carried out on the recent ASVspoof 2019 corpus, including both logical and physical access scenarios. The experimental results show that training a DNN based anti-spoofing system with our proposed loss functions clearly outperforms the performance of the same system being trained with other well-known loss functions. Moreover, the results also show that the proposed loss functions are effective for different types of neural network architectures.

On the study of replay and voice conversion attacks to text-dependent speaker verification

Automatic speaker verification (ASV) is to automatically accept or reject a claimed identity based on a speech sample. Recently, individual studies have confirmed the vulnerability of state-of-the-art text-independent ASV systems under replay, speech synthesis and voice conversion attacks on various databases. However, the behaviours of text-dependent ASV systems have not been systematically assessed in the face of various spoofing attacks. In this work, we first conduct a systematic analysis of text-dependent ASV systems to replay and voice conversion attacks using the same protocol and database, in particular the RSR2015 database which represents mobile device quality speech. We then analyse the interplay of voice conversion and speaker verification by linking the voice conversion objective evaluation measures with the speaker verification error rates to take a look at the vulnerabilities from the perspective of voice conversion.

On robustness of speech based biometric systems against voice conversion attack

Voice conversion (VC) approach, which morphs the voice of a source speaker to be perceived as spoken by a specified target speaker, can be intentionally used to deceive the speaker identification (SID) and speaker verification (SV) systems that use speech biometric. Voice conversion spoofing attacks to imitate a particular speaker pose potential threat to these kinds of systems. In this paper, we first present an experimental study to evaluate the robustness of such systems against voice conversion disguise. We use Gaussian mixture model (GMM) based SID systems, GMM with universal background model (GMM-UBM) based SV systems and GMM supervector with support vector machine (GMM-SVM) based SV systems for this. Voice conversion is conducted by using three different techniques: GMM based VC technique, weighted frequency warping (WFW) based conversion method and its variation, where energy correction is disabled (WFW−). Evaluation is done by using intra-gender and cross-gender voice conversions between fifty male and fifty female speakers taken from TIMIT database. The result is indicated by degradation in the percentage of correct identification (POC) score in SID systems and degradation in equal error rate (EER) in all SV systems. Experimental results show that the GMM-SVM SV systems are more resilient against voice conversion spoofing attacks than GMM-UBM SV systems and all SID and SV systems are most vulnerable towards GMM based conversion than WFW and WFW− based conversion. From the results, it can also be said that, in general terms, all SID and SV systems are slightly more robust to voices converted through cross-gender conversion than intra-gender conversion. This work extended the study to find out the relationship between VC objective score and SV system performance in CMU ARCTIC database, which is a parallel corpus. The results of this experiment show an approach on quantifying objective score of voice conversion that can be related to the ability to spoof an SV system.