Detection of Replay Attacks Using Single Frequency Filtering Cepstral Coefficients

Abstract

Automatic speaker verification systems are vulnerable to spoofing attacks. Recently, various countermeasures have been developed for detecting high technology attacks such as speech synthesis and voice conversion. However, there is a wide gap in dealing with replay attacks. In this paper, we propose a new feature for replay attack detection based on single frequency filtering (SFF), which provides high temporal and spectral resolution at each instant. Single frequency filtering cepstral coefficients (SFFCC) with Gaussian mixture model classifier are used for the experimentation on the standard BTAS-2016 corpus. The previously reported best result, which is based on constant Q cepstral coefficients (CQCC) achieved a half total error rate of 0.67 % on this data-set. Our proposed method outperforms the state of the art (CQCC) with a half total error rate of 0.0002 %.

IIIT-H Spoofing Countermeasures for Automatic Speaker Verification Spoofing and Countermeasures Challenge 2019

“…The spectrotemporal resolution of spectrum can be adjusted with the parameter r, which represents the pole location of a single pole filter in z-plane [17]. Recently, cepstral coefficients computed from low SNR instants of SFF spectrum are explored for replay attack detection [11,12]. In this study, we modified the SFFCC features proposed in [11] and named it as modified SFFCC (MSFFCC).…”

“…Recently, cepstral coefficients computed from low SNR instants of SFF spectrum are explored for replay attack detection [11,12]. In this study, we modified the SFFCC features proposed in [11] and named it as modified SFFCC (MSFFCC). As described in our previous work [12], MSFFCC will also contain three steps in its computation.…”

“…Successful countermeasures developed for SS and VC attacks are based on the following theories: (i) natural phase information is missing in the synthesized speech [9], (ii) the long term dynamics of speech are not properly modelled by the majority of the synthetic speech generating techniques [6], and (iii) the cues related to synthetic speech are spread across all frequency regions [7,10]. Successful countermeasures developed against replay attacks are based on the acoustic characterization of channel characteristics using high resolution features [6,11,12,13].…”

“…By considering the aforementioned fundamental ideas involved in the implementation of successful countermeasures for all the major spoof attacks (VC, SS and replay), in this study we investigated a set of instantaneous cepstral features. They are (a) single frequency cepstral coefficients (SFFCC) [11,12], (b) zero time windowing cepstral coefficients (ZTWCC) [13], and (c) instantaneous frequency cepstral coefficients (IFCC) [14,15] for both LA and PA tasks. The main motivation to use these features as spoofing countermeasures are as follows,…”

Deep Siamese Architecture Based Replay Detection for Secure Voice Biometric

“…Current countermeasures against replay attacks generally operate in one of three ways: (a) identifying exact reproduction of a previous access attempt; (b) exploiting differences in the speech transmission channel [9]; or (c) targeting artefacts in replayed speech such as pop-noise [10], and source features [11]. Commonly used spectral features include sub-band spectral centroid magnitude coefficients (SCMCs) [12], constant-Q cepstral coefficients (CQCCs) [13], single frequency filtering cepstral coefficients (SFF-CCs) [14], inverse Mel frequency cepstral coefficients (IMFCCs) [15], rectangular filter cepstral coefficients (RFCCs) [15], and scattering decomposition based features [16]. In addition, deep neural network (DNN) architectures have also been employed either as discriminative feature extractors [17] or as an end-toend spoofing detectors [18] in a number of ways.…”