Joint Speaker Verification and Antispoofing in the <inline-formula> <tex-math notation="LaTeX">$i$ </tex-math></inline-formula>-Vector Space

Abstract

Abstract-Any biometric recognizer is vulnerable to spoofing attacks and hence voice biometric, also called 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 considered as one of the most challenging for modern recognition systems. To detect spoofing, most existing countermeasures assume explicit or implicit knowledge of a particular VC system and focus on designing discriminative features. In this work, we explore backend generative models for more generalized countermeasures. Specifically, we model synthesis-channel subspace to perform speaker verification and anti-spoofing jointly in the i-vector space, which is a well-established technique for speaker modeling. It enables us to integrate speaker verification and anti-spoofing tasks into one system without any fusion techniques. To validate the proposed approach, we study vocoder-matched and vocodermismatched ASV and VC spoofing detection on the NIST 2006 speaker recognition evaluation dataset. Promising results are obtained for standalone countermeasures as well as their combination with ASV systems using score fusion and joint approach.

A broad review on non-intrusive active user authentication in biometrics

“…Devising effective techniques to select features with the highest relevance would alleviate this problem to a certain extent. The joint speaker verification design was discussed by Sizov et al (2015); it is another text-independent system that combines narrator verification and anti-spoofing by using an i-vector strategy for speaker modelling. The proficient score of the developed strategy is proven by incorporating the possible attacks.…”

“…Here, free text analysis of keystroke information was discussed, which allows continuous Here, the developed model has utilized the coefficients of mel-frequency to process the audio signal. Finally, the developed strategy has gained accuracy as 98.2% If the dataset is complex then the developed frame model has gained high error rate and this leads to fall in accuracy rate Moreover, the key reason for achieving high error percentage is in this model the coefficients of linear function is not defined Sizov et al (2015) Backend scheme It has gained the finest accuracy rate with short duration…”

Tandem Assessment of Spoofing Countermeasures and Automatic Speaker Verification: Fundamentals

“…Both the individual (4) and the tandem (7) error rates take values in [0, 1]. This is not the case for the t-DCF in (8), however, which is a linear combination of the tandem errors formed by non-negative but otherwise unconstrained multipliers (the products of costs and priors). The 'raw' t-DCF values can hence be difficult to interpret, especially across different t-DCF parametrizations.…”

“…Furthermore, CMs are always used in combination with ASV. Previous work has shown the potential to combine the action of ASV and CM systems in the form of a single, integrated system [8], by the back-end fusion of independently trained ASV and CM systems [9], [10], or via the tandem detection framework illustrated in Fig. 1.…”

Integrated Replay Spoofing-Aware Text-Independent Speaker Verification

“…Results demonstrate the vulnerability of speaker verification systems that are unaware of presentation attack detection (PAD). While a number of studies have worked to develop independent systems for SV and PAD, few have sought to integrate the SV and PAD systems [12][13][14][15][16][17]. More specifically, this handful of studies proposed approaches such as cascaded, parallel [12,13], and joint systems [14,16,17].…”