A new speaker verification spoofing countermeasure based on local binary patterns

Speech spoofing countermeasures based on source voice analysis and machine learning techniques

As speaker verification is widely used as a means of verifying personal identity in commercial applications, the study of antispoofing countermeasures has become increasingly important. By choosing appropriate spectral and prosodic feature mapping, spoofing methods based on voice conversion and speech synthesis are both capable of deceiving speaker verification systems that typically rely on these features. Consequently alternative front-ends are required for effective spoofing detection. This paper investigates the use of the recently proposed hierarchical scattering decomposition technique, which can be viewed as a generalization of all constant-Q spectral decompositions, to implement front-ends for stand-alone spoofing detection. The coefficients obtained using this decomposition are converted to a feature vector of Scattering Cepstral Coefficients (SCCs). We evaluate the performance of SCCs on the recent spoofing and Antispoofing (SAS) corpus as well as the ASVspoof 2015 challenge corpus and show that SCCs are superior to all other front-ends that have previously been benchmarked on the ASVspoof corpus.

Local binary pattern (LBP) is sensitive to image noise. Noise-resistant LBP (NRLBP) improves the robustness to noise by incorporating the prior knowledge of images and information of other LBP bits into encoding process. However, it encodes the small pixel difference in such a way that its sign and magnitude are ignored. Although the small pixel difference may be easily distorted by noise, some of its information is still useful for LBP encoding. In this letter, we propose two enhanced NRLBPs that jointly utilize the sign and the magnitude of the current pixel difference, and also the information of other LBP bits. The proposed approaches are validated on two benchmark databases and demonstrate a superior performance compared with NRLBP and other LBP variants. The performance gain is significant when the noise level is high.

This study focuses on a comprehensive study of Automatic Signature Verification (ASV) for off-line Thai signatures; an investigation was carried out to characterise the challenges in Thai ASV and to baseline the performance of Thai ASV employing baseline features, being Local Binary Pattern, Local Directional Pattern, Local Binary and Directional Patterns combined (LBDP), and the baseline shape/feature-based hidden Markov model. As there was no publicly available Thai signature database found in the literature, the authors have developed and proposed a database considering real-world signatures from Thailand. The authors have also identified their latent challenges and characterised Thai signature-based ASV. The database consists of 5,400 signatures from 100 signers. Thai signatures could be bi-script in nature, considering the fact that a single signature can contain only Thai or Roman characters or contain both Roman and Thai, which poses an interesting challenge for script-independent SV. Therefore, along with the baseline experiments, the investigation on the influence and nature of bi-script ASV was also conducted. From the equal error rates and Bhattacharyya distance, the score achieved in the experiments indicate that the Thai SV scenario is a script-independent problem. The open research area on this subject of research has also been addressed.

Any biometric recognizer is vulnerable to direct spoofing attacks and automatic speaker verification (ASV) is no exception; replay, synthesis and conversion attacks all provoke false acceptances unless countermeasures are used. We focus on voice conversion (VC) attacks. Most existing countermeasures use full knowledge of a particular VC system to detect spoofing. We study a potentially more universal approach involving generative modeling perspective. Specifically, we adopt standard ivector representation and probabilistic linear discriminant analysis (PLDA) back-end for joint operation of spoofing attack detector and ASV system. As a proof of concept, we study a vocoder-mismatched ASV and VC attack detection approach on the NIST 2006 speaker recognition evaluation corpus. We report stand-alone accuracy of both the ASV and countermeasure systems as well as their combination using score fusion and joint approach. The method holds promise. Index Terms: speaker recognition, spoofing, voice conversion attack, i-vector, joint verification and anti-spoofing

Automatic speaker verification systems play a significant role among current biometric authentication systems. However, they have become vulnerable to various types of malicious spoofing attacks with the increasing popularity. These attacks include identical twins, impersonation, speech synthesis, voice conversion and replay. Recent studies have highlighted the need for developing more effective and robust countermeasures for replay attacks, which can be very challenging to detect. However, replayed speech has shown some frequency band-based differences to non-replayed speech signals. This paper intends to capture these differences using multiple perceptual filter banks to extract a spectral centroid magnitude feature for replay attack detection. The paper further highlights the need for custom filter banks rather than using already available filter banks for enhancing the replay attack detection performance.

Existing Tagalog Text-to-speech (TTS) systems still have room for improvement, and although recent attempts at creating local TTS systems for Philippine spoken languages were able to generate synthesized speech, they still possess relatively low Mean Opinion Scores (MOS), ranging from 1.5 to 3.9 (out of 5), when it comes to naturalness and intelligibility. Improving speech prosody, the main factor for a speech's naturalness or individuality, has been made possible through voice conversion (VC). This project aims to implement a VC system for Tagalog synthesized speech, specifically using Cycle Generative Adversarial Networks (Cycle-GAN), a state-of-the-art neural network architecture used in non-parallel VC. Inter-gender and intra-gender VC were made for two types of inputs: Google's own Tagalog TTS and a locally sourced TTS system built from Mary TTS. Results show that Google TTS and its VC models perform better overall than Mary TTS and its VC models. Mel Cepstral Distortions (MCD) and F0: Root Mean Square Errors (F0:RMSE) vary across all models, reaching an MCD as low as 6.52 dB for Google TTS' intra-gender VC and an F0:RMSE as low as 16.92 Hz from Google TTS' inter-gender VC. Meanwhile, undergoing VC also caused a degradation in perceived speech quality as seen in a decrease in MOS across all VC models. Inter-gender VC for both TTS inputs were subjectively more preferred over intra-gender VC, reaching MOS values of 3.76 and 2.32 for Google TTS and Mary TTS inputs, respectively. Furthermore, it was also shown that male respondents were likely to rate higher opinion scores for intra-gender VC than female respondents, likely due to differences in hearing sensitivities.

In the last few years, we have witnessed an exponential growth in voice spoofing attacks. The intruders employ different types of attacks such as speech synthesis where they use the machine generated speech against any target person to fool the automatic speaker verification (ASV) systems for various tasks i.e. home control, bank account access, etc. The availability of modern-day advanced tools has made it convenient to launch such types of voice spoofing attacks. To overcome the challenges associated with bypassing the security of ASV systems using the synthetic speech, we propose an effective synthetic speech detector using a fusion of spectral features. More specifically, we propose a fused feature vector consisting of MFCC, GTCC, Spectral Flux, and Spectral Centroid for audio signal representation. This fused feature set is capable of capturing the traits of speech variation attributes of genuine signal and algorithmic artifacts of synthetic signals. These features are further used to train the bilstm to classify the signal as genuine or spoof. The proposed framework is capable of detecting both the voice conversion and synthetic speech attacks on ASV systems. Performance of our framework is evaluated on ASVspoof 2019 LA dataset. Our experimental results illustrate the effectiveness of the proposed framework for logical access attacks (voice conversion and cloned/synthetic voice) detection.

Well-known vulnerabilities of voice-based biometrics are impersonation, replay attacks, artificial signals/speech synthesis, and voice conversion. Among these, voice impersonation is the obvious and simplest way of attack that can be performed. Though voice impersonation by amateurs is considered not a severe threat to ASV systems, studies show that professional impersonators can successfully influence the performance of the voice-based biometrics system. In this work, we have created a novel voice impersonation attack dataset and studied the impact of voice impersonation on automatic speaker verification systems. The dataset consisting of celebrity speeches from 3 different languages, and their impersonations are acquired from YouTube. The vulnerability of speaker verification is observed among all three languages on both the classical i-vector based method and the deep neural network-based x-vector method.

Speaker recognition researchers acknowledge that systems which aim to verify speakers automatically based on their pronunciation of an utterance are vulnerable to spoofing attacks using voice conversion and speech synthesis technologies. The first automatic speaker verification spoofing and countermeasures challenge (ASVspoof2015) was designed to stimulate interest in this problem among the speaker recognition communities. In the course of the challenge and subsequently, it became clear that the most effective countermeasures against spoofing attacks are low-level acoustic features (typically extracted at 10 ms intervals) designed to detect artifacts in synthetic or voice converted speech. In this work, we demonstrate the effectiveness of the infinite impulse response - constant Q transform (IIR-CQT) spectrum-based cepstral coefficients (ICQC) as anti-spoofing front-end. The IIR-CQT spectrum is estimated by filtering the multi-resolution fast Fourier transform with an infinite impulse response filter. These features can be used on their own with a standard Gaussian mixture model backend to detect spoofing attacks or they can be used in tandem with bottleneck features which are extracted from a bottleneck layer in a deep neural network designed to discriminate between synthetic and natural speech. We show that the ICQC features are capable of producing very low equal error rates on the individual spoofing attacks in the ASVspoof2015 data set (0.02% on the known attacks, 0.23% on the unknown attacks, and 0.13% on average). Moreover, with a single decision threshold (common to all of the attacks), the ICQC front end yielded an equal error rate of 0.20%.

In conventional voice conversion methods, some features of a speech signal’s spectrum envelope are first extracted. Then, these features are converted so as to best match a target speaker’s speech by designing and using a set of conversions. Ultimately, the spectrum envelope of the target speaker’s speech signal is reconstructed from the converted features. The spectrum envelope reconstructed from the converted features usually deviates from its natural form. This aberration from the natural form observed in cases such as over-smoothing, over-fitting, and widening of formants is partially caused by two factors: (1) there is an error in the reconstruction of spectrum envelope from the features, and (2) the set of features extracted from the spectrum envelope of the speech signal is not closed. A method is put forward to improve the naturalness of speech by means of \(\epsilon \)-closed sets of extended vectors in voice conversion systems. In this approach, \(\epsilon \)-closed sets to reconstruct the natural spectrum envelope of a signal in the synthesis phase are introduced. The elements of these sets are generated by forming a group of extended vectors of features and applying a quantization scheme on the features of a speech signal. The use of this method in speech synthesis leads to a noticeable reduction of error in spectrum reconstruction from the features. Furthermore, the final spectrum envelope extracted from voice conversions maintains its natural form and, consequently, the problems arising from the deviation of voice from its natural state are resolved. The above method can be generally used as one phase of speech synthesis. It is independent of the voice conversion technique used and its parallel or non-parallel training method, and can be applied to improve the naturalness of the generated speech signal in all common voice conversion methods. Moreover, this method can be used in other fields of speech processing like texts to speech systems and vocoders to improve the quality of the output signal in the synthesis step.

This paper describes a technique for language-independent prosody modeling using unsupervised prosodic labelling in HMM-based speech synthesis and shows its applications to low bit-rate speech coding and speaker-independent voice conversion. In the proposed technique, sequences of prosodic features are roughly quantized at a phone level and the resultant indexes are used as the prosodic context for the model training. The conventional HMM-based speech synthesis requires accurate prosodic labels corresponding to the speech samples where manual modification is necessary to improve the modeling accuracy, which sometimes takes extra costs and limits its application. In contrast, the proposed technique creates the prosodic label from the training data itself and can apply not only to the speech synthesis but also to the speech coding and voice conversion. Subjective experimental results show the effectiveness of the use of the quantized F0 context without manual prosodic labelling.

An automatic speaker verification system aims to verify the speaker identity of a speech signal. However, a voice conversion system could manipulate a person’s speech signal to make it sound like another speaker’s voice and deceive the speaker verification system. Most countermeasures for voice conversion-based spoofing attacks are designed to discriminate bona fide speech from spoofed speech for speaker verification systems. In this paper, we investigate the problem of source speaker identification – inferring the identity of the source speaker given the voice converted speech. To perform source speaker identification, we simply add voice-converted speech data with the label of source speaker identity to the genuine speech dataset during speaker embedding network training. Experimental results show the feasibility of source speaker identification when training and testing with converted speeches from the same voice conversion model(s). In addition, our results demonstrate that having more converted utterances from various voice conversion model for training helps improve the source speaker identification performance on converted utterances from unseen voice conversion models.

Advances in species recognition technologies can contribute to the conservation and protection of flora species, especially those threatened with extinction. The aim of this research was to compare the early fusion approaches of operators known as Local Binary Patterns (LBP) and late fusion, carried out at the level of the decision classifiers, in the construction of an automatic recognition system of forest species. 1901 macroscopic images of wood from 46 Brazilian species were used. The extraction of image characteristics was done using two variants of the LBP descriptor, covering different aspects of spatial and angular resolution. The repeated stratified k-fold cross-validation method was used to estimate the performance of the classifiers. The cross-validation folds were created using stratified random sampling, whose strata were the prediction classes. An automatic recognition system based on the concatenation of rotation-invariant LBP histograms and the SVM classifier showed an F1-score of 97.67%. The fusion of classifiers, through majority voting, improved the F1-score of this system by 0.33% point. This experiment revealed that more than 50% of the species showed no misclassification or occurred only once or twice. It was identified that some groups of species generally confused by wood anatomists were perfectly differentiated by this classification system. The recognition system showed good ability to identify species, and if this technology is combined with traditional identification tools and empirical knowledge, it is possible to minimize errors in the identification of Brazilian flora, especially endangered species, for which the proposed classification system showed high accuracy.

Children's speaker verification in low and zero resource conditions