LP residual features to counter replay attacks
LP residual features to counter replay attacks
复制标题
用于对抗重放攻击的 LP 残余特征
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Debadatta Pati
中科院分区：
文献类型：
--
作者：
Jagabandhu Mishra;Madhusudan Singh;Debadatta Pati
重放攻击是一种利用目标预先录制的语音样本获取对自动说话人确认系统的非授权访问的方法。对策包括检测重放信号。主要线索包括跟踪记录和播放设备的特性，这些特性主要反映在由于扬声器的低频区域和由于两级A/D转换的高频区域。类脉冲线性预测（LP）残差信号的频谱分布在整个频率范围内，因此LP残差特征被证明对对抗重放攻击是有用的。根据Mel尺度在低频区域密集分布，在逆Mel尺度上反向分布的特点，利用剩余Mel频率倒谱系数（RMFCC）和剩余逆Mel频率倒谱系数（RIMFCC）特征进行对抗。这些功能的有效性在ASVspoof 2017数据库上得到了证明。对确定合适的预测阶数进行了初步研究。在等错误率（EER）方面，RMFCC功能提供了最好的性能，从20阶LP分析的14.75%和RIMFCC的16.20%，从10阶LP分析。RMFCC和RIMFCC特征的融合进一步提高了性能到10.40%，这是相对于最先进的谱质心幅度系数（SCMC）特征性能的11.49%。最后，RMFCC和RIMFCC特征与SCMC的融合提供了9.77%。这些结果证明了处理LP残差信号以对抗重放攻击的有用性。
Replay attack is a method of using targets prerecorded speech samples for acquiring unauthorized access to automatic speaker verification (SV) systems. The countermeasures involve detecting replay signals. The major clues include tracing the record and playback devices characteristics that dominantly reflect at low frequency regions due to loud speaker, and at high frequency regions due to two-stage A/D conversions. The spectral pattern of the impulse-like linear prediction (LP) residual signal is spread across entire frequency range, and so the LP residual features are conjectured to be useful to counter replay attacks. Based on the distribution nature of the mel-scale that tightly spaced in low frequency regions and reverse in inverse mel-scale, residual mel-frequency cepstral coefficients (RMFCC) and residual inverse mel-frequency cepstral coefficients (RIMFCC) features are used for countermeasure. The effectiveness of these features is demonstrated on ASVspoof2017 database. The initial study is made on deciding the suitable prediction order. In terms of equal error rate (EER), RMFCC features provide the best performance of 14.75% from 20th order LP analysis and RIMFCC of 16.20% from 10th order LP analysis. The fusion of RMFCC and RIMFCC features further improves the performance to 10.40%, that is comparatively better than the state-of-the-art spectral centroid magnitude co-efficients (SCMC) feature performance of 11.49%. Finally, the fusion of RMFCC and RIMFCC features together with SCMC provides 9.77%. These outcomes demonstrate the usefulness of processing LP residual signals to counter replay attacks.