End-To-End Audio Replay Attack Detection Using Deep Convolutional Networks with Attention
The DOI matches its Crossref record; 10 of the 10 references it cites were found in a registry or an index.
Abstract
With automatic speaker verification (ASV) systems becoming increasingly popular, the development of robust countermeasures against spoofing is needed. Replay attacks pose a significant threat to the reliability of ASV systems because of the relative difficulty in detecting replayed speech and the ease with which such attacks can be mounted. In this paper, we pro-pose an end-to-end deep learning framework for audio replay attack detection. Our proposed approach uses a novel visual attention mechanism on time-frequency representations of utterances based on group delay features, via deep residual learning (an adaptation of ResNet-18 architecture). Using a single model system, we achieve a perfect Equal Error Rate (EER) of 0% on both the development as well as the evaluation set of the ASVspoof 2017 dataset, against a previous best of 0.12% on the development set and 2.76% on the evaluation set reported in the literature. This highlights the efficacy of our feature representation and attention-based architecture in tackling the challenging task of audio replay attack detection.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“It means that the performance of the detection technique was really better than the previous ones; the best EER was 12% in the development set and 2.76% on the evaluation as stated in other literature (Tom et al., 2018).”
From Audio deepfakes: A survey (2023) · cites this paper as Tom et al., 2018
In the citing paper
“Additionally, before the ASVspoof Challenge 2017 (Kinnunen et al., 2017; Lavrentyeva et al., 2017), there were only a couple of research papers done on replay attack, and after this challenge, more approaches for this attack were researched (Tom et al., 2018; Pradhan et al., 2019).”
From Audio deepfakes: A survey (2023) · cites this paper as Tom et al., 2018
In the citing paper
“A current technique that detects end-to-end replay attacks is by using deep convolutional networks (Tom et al., 2018), which is shown in Figure 3.”
From Audio deepfakes: A survey (2023) · cites this paper as Tom et al., 2018
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 5×
“In our work, we use the ResNet-18 model [9], which consists of layers in the following order: 7 × 7 convolution layer, eight residual blocks, Global Average Pooling (GAP) layer, followed by a fully connected layer with softmax.”
BackgroundMethodInfluentialLearning Deep Features for Discriminative Localization
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.319 · Cited by 11,428 · cited 2×
“To capture the discriminative information contained within specific regions, we propose a visual attention mechanism based on class activation mapping [17] for replay attack detection.”
BackgroundMethodThe ASVspoof 2017 challenge: assessing the limits of replay spoofing attack detection
Kinnunen, T., Sahidullah, M., Delgado, H., Todisco, M., Evans, N., Yamagishi, J., & Lee, K. A. (2017) · Interspeech 2017 · doi:10.21437/interspeech.2017-1111 · Cited by 535 · cited 4×
“Only 20 of those 49 submissions outperformed the baseline spoof detection system, which was based on a Gaussian mixture model (GMM) back-end classifier with constant Q cepstral co-efficient (CQCC) features [5].”
BackgroundInfluentialAudio replay attack detection with deep learning frameworks
Lavrentyeva, G., Novoselov, S., Malykh, E., Kozlov, A., Kudashev, O., & Shchemelinin, V. (2017) · Interspeech 2017 · doi:10.21437/interspeech.2017-360 · Cited by 292 · cited 1×
“Attention Mask Replay SVM i-vect , CNN FFT +RNN [21] 3.95% - 6.73% - VESA-IFCC + CFCCIF [22] 0.12% - 18.33% - Grand Ensemble [8] - - 2.76% - ASVspoof 2017 baseline [5] - - 24.77% - GD-ResNet-18 w/ Attention 0.0% 0.0% 0.0% 0.0% Table 2 : Results on the ASVspoof 2017 dataset”
InfluentialToward a Universal Synthetic Speech Spoofing Detection Using Phase Information
Sanchez, J., Saratxaga, I., Hernaez, I., Navas, E., Erro, D., & Raitio, T. (2015) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2015.2398812 · Cited by 88 · cited 1×
“Previously, phase information has been used for detecting speech synthesis and voice conversion attacks [14].”
MethodRe-assessing the threat of replay spoofing attacks against automatic speaker verification
Alegre, F., Janicki, A., & Evans, N. (2014) · Graduate School and Research Center in Digital Science (EURECOM) · Cited by 80
Patil, H. A., Kamble, M. R., Patel, T. B., & Soni, M. H. (2017) · Interspeech (USB) · doi:10.21437/interspeech.2017-1362 · Cited by 77 · cited 1×
“Attention Mask Replay SVM i-vect , CNN FFT +RNN [21] 3.95% - 6.73% - VESA-IFCC + CFCCIF [22] 0.12% - 18.33% - Grand Ensemble [8] - - 2.76% - ASVspoof 2017 baseline [5] - - 24.77% - GD-ResNet-18 w/ Attention 0.0% 0.0% 0.0% 0.0% Table 2 : Results on the ASVspoof 2017 dataset”
InfluentialApplication of automatic speaker recognition techniques to pathological voice assessment (dysphonia)
Fredouille, C., Pouchoulin, G., Bonastre, J.-F., Azzarello, M., Giovanni, A., & Ghio, A. (2005) · Interspeech (USB) · doi:10.21437/interspeech.2005-90 · Cited by 52 · cited 1×
“Automatic speaker verification (ASV) technology has several applications, including voice-based identification, pathological voice assessment [1] and forensic evidence evaluation [2].”
BackgroundLS regularization of group delay features for speaker recognition
Kua, J. M. K., Epps, J., Eliathamby Ambikairajah, A., & Choi, E. H. C. (2009) · Interspeech (USB) · doi:10.21437/interspeech.2009-46 · Cited by 22 · cited 1×
“More-over, GD-gram contains both power and phase spectrum information [11, 13], thus making it a good feature representation for end-to-end learning for spoof detection.”
BackgroundApplication of automatic speaker verification techniques for forensic evidence evaluation
Adikari, A. M. T. S. B., Devadithya, S., Bandara, A. R. S. T., Dharmawardane, K. C. J., & Wavegedara, C. B. (2014) · IEEE International Conference on Digital Signal Processing (DSP) · doi:10.1109/icdsp.2014.6900703 · Cited by 3 · cited 1×
“Automatic speaker verification (ASV) technology has several applications, including voice-based identification, pathological voice assessment [1] and forensic evidence evaluation [2].”
Background
Cited by
Deepfake Audio Detection via MFCC features using Machine Learning
Hamza, A., Javed, A. R. R., Iqbal, F., Kryvinska, N., Almadhor, A. S., Jalil, Z., & Borghol, R. (2022) · IEEE Access · doi:10.1109/access.2022.3231480 · Cited by 264
“As of now, deep convolutional networks are used as a method for detecting complete replay attacks [22].”
MethodRecurrent convolutional structures for audio spoof and video deepfake detection
Chintha, A., Thai, B., Sohrawardi, S. J., Bhatt, K., Hickerson, A., Wright, M., & Ptucha, R. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.2999185 · Cited by 226
Nautsch, A., Wang, X., Evans, N., Kinnunen, T. H., Vestman, V., Todisco, M., Delgado, H., Sahidullah, M., Yamagishi, J., & Lee, K. A. (2021) · IEEE Transactions on Biometrics, Behavior, and Identity Science · doi:10.1109/tbiom.2021.3059479 · Cited by 185
“T10 [42]: Group delay (GD) grams [43] with cepstral mean and variance normalization (CMVN) with data augmentation via speed pertubation.”
MethodKhanjani, Z., Watson, G., & Janeja, V. P. (2023) · Frontiers in Big Data · doi:10.3389/fdata.2022.1001063 · Cited by 121
“A current technique that detects end-to-end replay attacks is by using deep convolutional networks (Tom et al., 2018), which is shown in Figure 3. is an overview of the end-to-end replay attacks’ detection framework by using deep convolutional networks (GD: Group Delay, GAP: Global Average Pooling,…”
BackgroundMethodInfluentialAdvances in anti-spoofing: from the perspective of ASVspoof challenges
Kamble, M. R., Sailor, H. B., Patil, H. A., & Li, H. (2020) · APSIPA Transactions on Signal and Information Processing · doi:10.1017/atsip.2019.21 · Cited by 102
A Comprehensive Review of DeepFake Detection Using Advanced Machine Learning and Fusion Methods
Gupta, G., Raja, K., Gupta, M., Jan, T., Whiteside, S. T., & Prasad, M. (2023) · Electronics · doi:10.3390/electronics13010095 · Cited by 99
“Tom et al. [50] proposed a detection method consisting of a group delay gram (GD-gram) obtained by concatenating a group delay function over consecutive frames as a novel time-frequency representation of an utterance.”
MethodCai, W., Wu, H., Cai, D., & Li, M. (2019) · Interspeech 2019 · doi:10.21437/interspeech.2019-1230 · Cited by 65
A Gated Recurrent Convolutional Neural Network for Robust Spoofing Detection
Gomez-Alanis, A., Peinado, A. M., Gonzalez, J. A., & Gomez, A. M. (2019) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2019.2937413 · Cited by 53
“Compared to other relevant systems of the literature, our proposal is only outperfomed by the GD-ResNet-18 [57] which obtains a perfect EER of 0% on the evaluation set of the version 1 database.”
MethodModified Magnitude-Phase Spectrum Information for Spoofing Detection
Yang, J., Wang, H., Das, R. K., & Qian, Y. (2021) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2021.3060810 · Cited by 53
Jung, J.-W., Shim, H.-J., Heo, H.-S., & Yu, H.-J. (2019) · Interspeech (USB) · doi:10.21437/interspeech.2019-1991 · Cited by 65
A Survey on Replay Attack Detection for Automatic Speaker Verification (ASV) System
Patil, H. A., & Kamble, M. R. (2018) · 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipa.2018.8659666 · Cited by 48
Gupta, S., Maple, C. R., Crispo, B., Raja, K. B., Yautsiukhin, A., & Martinelli, F. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.109453 · Cited by 64
Leveraging edges and optical flow on faces for deepfake detection
Chintha, A., Rao, A. G., Sohrawardi, S. J., Bhatt, K., Wright, M., & Ptucha, R. (2020) · IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb48548.2020.9304936 · Cited by 54
“Detection methods target either high-level bio-metric features [4, 2, 40, 15] or low-level structural features [1, 63, 24, 50, 56, 16, 49] with separate audio spoof detection methods [32, 70, 34].”
MethodReplay detection using CQT-based modified group delay feature and ResNeWt network in ASVspoof 2019
Cheng, X., Xu, M., & Zheng, T. F. (2019) · 2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc47483.2019.9023158 · Cited by 37
“ResNet is a popular model in image recognition [19], also has been utilized in the replay detection task [10], [13], [16].”
MethodLow frequency frame-wise normalization over constant-Q transform for playback speech detection
Yang, J., & Das, R. K. (2019) · Digital Signal Processing · doi:10.1016/j.dsp.2019.02.018 · Cited by 39
Dataset artefacts in anti-spoofing systems: A case study on the ASVspoof 2017 benchmark
Chettri, B., Benetos, E., & Sturm, B. L. T. (2020) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2020.3036777 · Cited by 35
“Furthermore, attention-based models for replay detection have also been studied in [26] and [22].”
BackgroundMethodOpen Challenges in Synthetic Speech Detection
Cuccovillo, L., Papastergiopoulos, C., Vafeiadis, A., Yaroshchuk, A., Aichroth, P., Votis, K., & Tzovaras, D. (2022) · 2022 IEEE International Workshop on Information Forensics and Security (WIFS) · doi:10.1109/wifs55849.2022.9975433 · Cited by 33
“The ability of deep neural networks to extract meaningful features from the raw input has resulted in building end-to-end systems capable of recognizing synthetic audio [26] and replay attacks [27].”
BackgroundMel-Spectrogram Image-Based End-to-End Audio Deepfake Detection Under Channel-Mismatched Conditions
Fathan, A., Alam, J., & Kang, W. H. (2022) · 2022 IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme52920.2022.9859621 · Cited by 29
“…both LA and PA tasks [16], only a handful of acoustic features such as Constant Q Cepstral Coefficient (CQCC) [17], Inverted Mel-Frequency Cepstral Coefficients (IMFCC), Joint Gram [18], Group Delay Gram (GD Gram) [19], and RawNet2 [20] have been proved to be useful for anti-spoofing task.”
BackgroundMulti-Scale Permutation Entropy for Audio Deepfake Detection
Wang, C., He, J., Yi, J., Tao, J., Zhang, C. Y., & Zhang, X. (2024) · ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp48485.2024.10448095 · Cited by 31
“…CQCC, ∗ First Author and Second Author contribute equally to this work. several other feature sets also exhibit strong performance in anti-spoofing tasks, including the Group Delay Gram [5], Log Power Spectrum (LPS) [6], and Cochlear Filter Cepstral Coefficients In-stantaneous Frequency (CFCCIF) [7].”
BackgroundA New Replay Attack Against Automatic Speaker Verification Systems
Yoon, S.-H., Koh, M.-S., Park, J., & Yu, H.-J. (2020) · IEEE Access · doi:10.1109/access.2020.2974290 · Cited by 34
“The attention mechanism applied in [34] should be supported by high classification accuracy.”
BackgroundMethodInfluentialNew Acoustic Features for Synthetic and Replay Spoofing Attack Detection
Wei, L., Long, Y., Wei, H., & Li, Y. (2022) · Symmetry · doi:10.3390/sym14020274 · Cited by 29
“For instance, the convolutional neural network (CNN) was used to learn deep features from the group delay [25] and Siamese embedding the spectrogram [26].”
MethodReplay Attack Detection Using Linear Prediction Analysis-Based Relative Phase Features
Phapatanaburi, K., Wang, L., Nakagawa, S., & Iwahashi, M. (2019) · IEEE Access · doi:10.1109/access.2019.2960369 · Cited by 25
Kwak, I., Kwag, S., Lee, J., Jeon, Y., Hwang, J., Choi, H.-J., Yang, J., Han, S.-Y., Huh, J. H., Lee, C.-H., et al. (2023) · IEEE Access · doi:10.1109/access.2023.3275790 · Cited by 30
“When tested against the ASVspoof 2017 and ASVspoof 2019 evaluation datasets, both LCNN and ResNet-based models demonstrated excellent results with regard to EERs [8], [9], [13], [14].”
MethodHow Deep Are the Fakes? Focusing on Audio Deepfake: A SurveyPreprint
Khanjani, Z., Watson, G., & Janeja, V. P. (2021) · arXiv (Cornell University) · doi:10.13016/m2ivas-ogyu · Cited by 22
ResMax: Detecting Voice Spoofing Attacks with Residual Network and Max Feature Map
Kwak, I., Kwag, S., Lee, J., Huh, J. H., Lee, C.-H., Jeon, Y., Hwang, J., & Yoon, J. W. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9412165 · Cited by 27
“Both LCNN and ResNet based models have performed well against the ASVspoof 2017 and ASVspoof 2019 evaluation datasets, demonstrating top performances with respect to EERs [6], [7], [9], [10].”
MethodSignificance of Constant-Q Transform for Voice Liveness Detection
Khoria, K., Patil, A. T., & Patil, H. A. (2021) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco54536.2021.9615993 · Cited by 19
Improving anti-spoofing with octave spectrum and short-term spectral statistics information
Yang, J., & Das, R. K. (2019) · Applied Acoustics · doi:10.1016/j.apacoust.2019.107017 · Cited by 18
Spoof Detection Using Time-Delay Shallow Neural Network and Feature Switching
Kumar, M. G., Kumar, S. R., S, S. M., Bharathi, B., & Murthy, H. A. (2019) · IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU) · doi:10.1109/asru46091.2019.9003824 · Cited by 20
“In [7] an end-to-end neural network (NN) with attention masking was proposed to learn the difference in the spectrogram of bonafide and the replayed utterances.”
BackgroundApproaches for Fake Content Detection: Strengths and Weaknesses to Adversarial Attacks
Carter, M., Tsikerdekis, M., & Zeadally, S. (2020) · IEEE Internet Computing · doi:10.1109/mic.2020.3032323 · Cited by 23
“20 One approach 21 for detecting fake audio focuses on detecting audio samples where the intonation has been intentionally faked.”
BackgroundSource Tracing: Detecting Voice Spoofing
Zhu, T., Wang, X., Qin, X., & Li, M. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9980129 · Cited by 32
Single and Multi-Speaker Cloned Voice Detection: From Perceptual to Learned Features
Barrington, S., Barua, R., Koorma, G., & Farid, H. (2023) · 2023 IEEE International Workshop on Information Forensics and Security (WIFS) · doi:10.1109/wifs58808.2023.10374911 · Cited by 17
“And, although somewhat outside of the scope of our work, there has also been an effort to detect audio spoofing in the form of a rebroadcast attack in which a person’s voice is recorded and replayed [24], [25].”
BackgroundWang, L., Yu, L., Zhang, Y., & Xie, H. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3414340 · Cited by 25
“On the one hand, some previous works focus on finding better front-end speech features, such as constant Q cepstral coefficients (CQCC) [7], single frequency filtering cepstral coefficients (SFFCC) [11] and group delay grams (GD-grams) [12].”
BackgroundVoice Presentation Attack Detection through Text-Converted Voice Command Analysis
Kwak, I., Huh, J. H., Han, S. T., Kim, I., & Yoon, J. (2019) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300828 · Cited by 16
“Voice replay attacks (replaying device owners’ recorded voices through a speaker) [8, 18, 32, 36] and voice synthesis attacks (new voice commands are generated using device owners’ voice samples and trained models, and played through a speaker) [25, 27, 35] are also effective against voice…”
BackgroundReplay attack detection using variable-frequency resolution phase and magnitude features
Liu, M., Wang, L., Dang, J., Lee, K. A., & Nakagawa, S. (2020) · Computer Speech & Language · doi:10.1016/j.csl.2020.101161 · Cited by 17
Transmission Line Cochlear Model Based AM-FM Features for Replay Attack Detection
Gunendradasan, T., Irtza, S., Ambikairajah, E., & Epps, J. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8682771 · Cited by 18
“convolutional neural networks and Siamese network have been proposed which are effective in discriminating replay attacks [12], [13], [14].”
BackgroundReplay Attack Detection Using Magnitude and Phase Information with Attention-based Adaptive Filters
Liu, M., Wang, L., Dang, J., Nakagawa, S., Guan, H., & Li, X. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8682739 · Cited by 16
“As the bestperforming system it implies promising effectiveness of the feature containing phase for this task [10].”
BackgroundMewada, H. K., Al‐Asad, J. F., Almalki, F. A., Khan, A. H., Almujally, N. A., El-Nakla, S., & Naith, Q. H. (2023) · Sensors · doi:10.3390/s23146637 · Cited by 20
“In Reference [48], a group delay concatenated over the consecutive frames of the speech signal was used as a feature in the ResNET18 classifier.”
MethodXu, L., Yang, J., You, C. H., Qian, X., & Huang, D. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3267610 · Cited by 19
“As DNN and ResNet are often selected for replay speech detection with promising performance in the previous work [31], [36], [37], [52], [53], [54], they are selected as classifiers in this article.”
MethodFake audio detection using Hierarchical Representations Learning and Spectrogram Features
Ali, M., Sabir, A., & Hassan, M. (2021) · International Conference on Robotics and Automation in Industry (ICRAI) · doi:10.1109/icrai54018.2021.9651401 · Cited by 14
Fast and Lightweight Voice Replay Attack Detection via Time-Frequency Spectrum Difference
He, R., Cheng, Y., Zheng, Z., Ji, X., & Xu, W. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3406962 · Cited by 17
“Tom et al. [17] used a residual network (ResNet) as a detection model, achieving 0% EER on the ASVspoof test set.”
MethodVoShield: Voice Liveness Detection with Sound Field Dynamics
Yang, Q., Cui, K., & Zheng, Y. (2023) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom53939.2023.10229038 · Cited by 13
“To defend against such attacks, existing works enhance ASV systems with liveness detection.”
MethodInfluentialReplay anti-spoofing countermeasure based on data augmentation with post selection
Zhao, Y., Togneri, R. B., & Sreeram, V. (2020) · Computer Speech & Language · doi:10.1016/j.csl.2020.101115 · Cited by 15
Deep learning countermeasures for detecting replay speech attacks: a review
Veesa, S., & Singh, M. (2024) · International Journal of Speech Technology · doi:10.1007/s10772-024-10163-z · Cited by 16
“The notable performance of proposed approaches in studies (Lavrentyeva et al., 2017; Sriskandaraja et al., 2018; Tom et al., 2018; Chen et al., 2018) on the ASVspoof 2017 database reveals the strength of deep learning frameworks from replay speech detection perspectives.”
InfluentialPhoneme Specific Modelling and Scoring Techniques for Anti Spoofing System
Suthokumar, G., Sriskandaraja, K., Sethu, V., Wijenayake, C. A., & Eliathamby Ambikairajah, A. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8682411 · Cited by 11
Speech frame selection for spoofing detection with an application to partially spoofed audio-data
Kumar, A. K., Paul, D., Pal, M., Sahidullah, M., & Saha, G. (2021) · International Journal of Speech Technology · doi:10.1007/s10772-020-09785-w · Cited by 14
“A DNN trained with attention mechanism has helped to accurately detect replay-based spoofed voice in version 1.0 of the ASVspoof 2017 dataset (Tom et al. 2018) 1 .”
InfluentialDeep correlation network for synthetic speech detection
Chen, C., Dai, B., Bai, B., & Chen, D. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.111413 · Cited by 12
DNN controlled adaptive front-end for replay attack detection systems
Wickramasinghe, B., Eliathamby Ambikairajah, A., Sethu, V., Epps, J., Li, H., & Dang, T. (2023) · Speech Communication · doi:10.1016/j.specom.2023.102973 · Cited by 13
Discriminative features based on modified log magnitude spectrum for playback speech detection
Yang, J., Xu, L., Ren, B., & Ji, Y. (2020) · EURASIP Journal on Audio Speech and Music Processing · doi:10.1186/s13636-020-00173-5 · Cited by 12
A Practical Guide to Logical Access Voice Presentation Attack Detection
Wang, X., & Yamagishi, J. (2022) · Studies in autonomic, data-driven and industrial computing · doi:10.1007/978-981-19-1524-6_8 · Cited by 24
Device Feature Extraction Based on Parallel Neural Network Training for Replay Spoofing Detection
You, C. H., & Yang, J. (2020) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2020.3011320 · Cited by 11
“In back-end classification part, most models for the spoofing detection are based on GMM [40] [41] [42], linear discriminative analysis (LDA) [35], support vector machine (SVM) [34] [43] [44], residual network (ResNet) [27], DNN [23] [24] and BLSTM [45] [46].”
BackgroundMethodReplay attack detection with raw audio waves and deep learning framework
Shukla, S., Prakash, J., & Guntur, R. S. (2019) · International Conference on Data Science and Engineering · doi:10.1109/icdse47409.2019.8971793 · Cited by 10
A multi-branch ResNet with discriminative features for detection of replay speech signals
Cheng, X., Xu, M., & Zheng, T. F. (2020) · APSIPA Transactions on Signal and Information Processing · doi:10.1017/atsip.2020.26 · Cited by 11
“[18] pre-trained a ResNet model in the ImageNet dataset [24] and then transformed it for replay detection tasks.”
MethodAvila, A. R., Alam, J., Prado, F., O’Shaughnessy, D. D., & Falk, T. H. (2020) · Computer Speech & Language · doi:10.1016/j.csl.2020.101163 · Cited by 9
Auditory Inspired Spatial Differentiation for Replay Spoofing Attack Detection
Wickramasinghe, B., Ambikairajah, E., Epps, J., Sethu, V., & Li, H. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8683693 · Cited by 9
Multi-Task Siamese Neural Network for Improving Replay Attack Detection
von Platen, P., Tao, F., & Tür, G. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-86 · Cited by 10
Adversarial Multi-Task Learning for Speaker Normalization in Replay Detection
Suthokumar, G., Sethu, V., Sriskandaraja, K., & Eliathamby Ambikairajah, A. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9054322 · Cited by 10
Attacks on Voice Assistant Systems
Lien, J., Momin, M. A. A., & Yuan, X. (2021) · Advances in web technologies and engineering book series · doi:10.4018/978-1-7998-7323-5.ch004 · Cited by 7
Modified Group Delay Cepstral Coefficients for Voice Liveness Detection
Singh, S., Khoria, K., & Patil, H. A. (2021) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco54536.2021.9616343 · Cited by 7
Mankad, S. H., & Garg, S. (2020) · Progress in Artificial Intelligence · doi:10.1007/s13748-020-00216-0 · Cited by 7
“More-over, when this signal is played back to the system while attempting the attack, it gets additional characteristics of the playback device [45–48].”
BackgroundThe Audio Auditor: User-Level Membership Inference in Internet of Things Voice Services
Miao, Y., Xue, M., Chen, C., Pan, L., Zhang, J., Zhao, B. Z. H., Kaafar, D., & Xiang, Y. (2020) · Proceedings on Privacy Enhancing Technologies · doi:10.2478/popets-2021-0012 · Cited by 49
“Recent works [18] and [44] propose an audio cloning attack and audio reply attack against the speech recognition system to impersonate a legitimate user or inject unintended voices.”
BackgroundEfficientWord-Net: An Open Source Hotword Detection Engine Based on Few-Shot Learning
Chidhambararajan, R., Rangapur, A., Chakkaravarthy, S. S., Cherukuri, A. K., Cruz, M. V., & Ilango, S. S. (2022) · Journal of Information & Knowledge Management · doi:10.1142/s0219649222500599 · Cited by 10
“The problem of detecting hotwords from audio streams started ever since the advent of voice-enabled IoT devices [15, 14, 17, 24, 25, 5, 29, 13, 3, 26, 22].”
BackgroundRoom-Scale Voice Liveness Detection for Smart Devices
Yang, Q., Cui, K., & Zheng, Y. (2024) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2024.3367269 · Cited by 6
“To ensure the security of voice assistants, automatic speaker verification (ASV) based on voiceprints [2], [3] is typically used to authenticate legitimate users.”
Anti-Replay: A Fast and Lightweight Voice Replay Attack Detection System
Shi, Z., Li, C., Jin, Z., Sun, W., Ji, X., & Xu, W. (2021) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads53394.2021.00027 · Cited by 6
“Second, software-based detection methods mainly rely on the inherently feature difference between the real human voice and † Corresponding author. the replayed audio, including acoustic features [5]–[7] and pop noise [8].”
MethodDeepfake Audio Detection: A Comparative Study of Advanced Deep Learning Models
Verma, K., Mittal, D., Samanta, S., Gulati, K., Kulkarni, O., Dar, M. A., & Biji, C. L. (2025) · IEEE Access · doi:10.1109/access.2025.3611839 · Cited by 12
“The authors [14] proposed an approach that integrates deep residual learning, specifically an adaptation of the ResNet-18 architecture.”
Multiresolution Decomposition Analysis via Wavelet Transforms for Audio Deepfake Detection
Fathan, A., Alam, J., & Kang, W. H. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20980-2_17 · Cited by 5
Fooling the Forgers: A Multi-Stage Framework for Audio Deepfake Detection
Kashyap, G. S., Siddiqui, Z. H., Azeez, M. A., Ali, R., Kumar, S., Kamuni, N., & Gao, J. (2025) · ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp49660.2025.10888175 · Cited by 4
“As an example, [6] constructed a network to use CNN for audio deepfake classification and revealed satisfactory efficacy on standard datasets.”
Dynamically Mitigating Data Discrepancy with Balanced Focal Loss for Replay Attack Detection
Dou, Y., Yang, H., Yang, M., Xu, Y., & Ke, D. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9412749 · Cited by 4
Detecting Replay Attacks Using Single-Channel Audio: The Temporal Autocorrelation of Speech
Lee, S.-K., Tsao, Y., & Wang, H. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9980046 · Cited by 4
“For spoofed speech detection, group delay [5], [6] and product spectrum [14] show excellent performance in replay attack detection.”
BackgroundMethodInfluentialUsing Multi-Resolution Feature Maps with Convolutional Neural Networks for Anti-Spoofing in ASV
Wang, Q., Lee, K. A., & Koshinaka, T. (2020) · The Speaker and Language Recognition Workshop (Odyssey) · doi:10.21437/odyssey.2020-20 · Cited by 4
Attention Network with GMM Based Feature for ASV Spoofing Detection
Lei, Z., Yu, H., Yang, Y., & Ma, M. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-86608-2_50 · Cited by 3
Wang, Y., Wang, X., Nishizaki, H., & Li, M. (2022) · International Symposium on Chinese Spoken Language Processing (ISCSLP) · doi:10.1109/iscslp57327.2022.10038240 · Cited by 6
“Through this series of challenges, the ASV anti-spoofing field has established two main anti-spoofing countermeasure research scenarios and databases: logical access (LA) considers spoofing attacks from text-to-speech synthesis (TTS) and voice conversion (VC), physical access (PA) refers to attacks produced by recordi…”
BackgroundSpoofed Speech Detection with Weighted Phase Features and Convolutional Networks
Dişken, G. (2022) · Archives of Acoustics · doi:10.24425/aoa.2022.141648 · Cited by 2
“In ( Tom et al ., 2018), group delay features and convolutional networks with attention are used to achieve 0.0% EER value for the ASVspoof 2017 database version 1.0.”
MethodInfluentialAn analysis of speaker dependent models in replay detection
Suthokumar, G., Sriskandaraja, K., Sethu, V., Eliathamby Ambikairajah, A., & Li, H. (2020) · APSIPA Transactions on Signal and Information Processing · doi:10.1017/atsip.2020.9 · Cited by 7
“Attention mechanisms [29, 30] and residual networks [31] have shown the most promise for replay detection because they help to emphasize salient regions of the input and help the network to generalize well for the small amount of training data.”
BackgroundSpoof detection using x-vector and feature switchingPreprint
Kumar, M. G., Kumar, S. R., S, S. M., Bharathi, B., & Murthy, H. A. (2019) · arXiv (Cornell University) · Cited by 2
A study on the role of subsidiary information in replay attack spoofing detectionPreprint
Jung, J.-W., Shim, H.-J., Heo, H.-S., & Yu, H.-J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2001.11688 · Cited by 3
“End-to-end (E2E) DNN is a promising architecture for various tasks [2, 25, 26].”
BackgroundEfficientWord-Net: An Open Source Hotword Detection Engine based on One-shot LearningPreprint
Chidhambararajan, R., Rangapur, A., & Sethuraman, S. C. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.00379 · Cited by 2
“The problem of detecting hotwords from audio streams started ever since the advent of voice-enabled IoT devices [15, 14, 17, 24, 25, 5, 29, 13, 3, 26, 22].”
BackgroundChaiwongyen, A., Pinkeaw, K., Kongprawechnon, W., Karnjana, J., & Unoki, M. (2021) · International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP) · doi:10.1109/isai-nlp54397.2021.9678164 · Cited by 2
Gupta, S. K. S., Maple, C. R., Crispo, B., Raja, K. B., Yautsiukhin, A., & Martinelli, F. (2022) · doi:10.36227/techrxiv.20115515 · Cited by 1
Replay Attack Detection Based on Spatial and Spectral Features of Stereo Signal
Yaguchi, R., Shiota, S., Ono, N., & Kiya, H. (2021) · Journal of Information Processing · doi:10.2197/ipsjjip.29.275 · Cited by 3
“For input features, spectrogram and phase information [22], [27], linear frequency cepstral coe ffi cients (LFCC) [18], constant Q cepstral coe ffi cients (CQCC) [17], Mel-frequency cepstral coe ffi cients (MFCC), inverted MFCC (IMFCC) [28], and rectangular filter cepstral co-e ffi cients (RFCC) [29] were…”
BackgroundSpectral Root Features for Replay Spoof Detection in Voice Assistants
Patil, A. T., Kotta, H., Acharya, R., & Patil, H. A. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-87802-3_46
Gupta, S., Maple, C. R., Crispo, B., Raja, K. B., Yautsiukhin, A., & Martinelli, F. (2022) · doi:10.36227/techrxiv.20115515.v1
EPTLENet: Replay Attack Detection with Efficient Parameter Transfer Learning Based on ERes2Net
Qian, Q., Kuang, Y.-L., & Yue, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-981-97-5606-3_35
$\beta$-Order Energy-Weighting Modulation on Spectral Bins for Replay Speech Detection
Yang, J., You, C. H., & Das, R. K. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3587454
“The features for RSD can be broadly categorized into two types: deep features [20], [21], [22] and hand-crafted features [23], [24], [25], [26], [27], [28].”
Erlacher, L. (2025) · ACM SIGSAC Conference on Computer and Communications Security (CCS) · doi:10.1145/3719027.3765567 · Cited by 1
“Replay attack detection has been studied using convolutional and recurrent neural models with specialized inputs, such as MFCCs, Gammatone features, or group-delay representations [2, 10, 17].”
Deep Learning for Audio-Visual Deepfake Detection: A Study on Face and Voice Synthesis
Harshitha, B., Hasmitha Krishna, N., Jayakaran, & Kannadasan, R. (2025) · International Conference on Electronics, Communication and Aerospace Technology (ICECA) · doi:10.1109/iceca66444.2025.11383505
Deepsheild: Efficient and Scalable Web-Based Deepfake Detection for Audio and Video Content
Newaskar, D., Marathe, S., Patil, D. S., Andhare, A., Pillai, A., Jadhav, T., & Sonawane, D. A. (2026) · International Conference on Emerging Technologies and Future Innovations (ETFI) · doi:10.1109/etfi68128.2026.11484683
MSLAT: multi-stream lip-aware transformer for temporal forgery localization
Li, W. A. N. G., Ziguang, L. I., Lingyun, Y. U., Hongtao, X. I. E., & ZHANG, Y. (2026) · Scientia Sinica Informationis · doi:10.1360/ssi-2025-0323
Metadata
- Authors
- Tom, F., Jain, M., & Dey, P.
- Venue
- Interspeech 2018
- Publisher
- ISCA
- Published
- 2018-09-02
- Pages
- 681-685
- ISSN
- 2308-457X (ISSN-L)
- OpenAlex
- W2889361425
- Cited by
- 89
- References
- 10
- Citations by year
- 2026: 2 · 2025: 5 · 2024: 7 · 2023: 9 · 2022: 13 · 2021: 17 · 2020: 20 · 2019: 15 · 2018: 1
- FWCI
- 5.39
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Interspeech (USB) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Music and Audio Processing
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Keywords
- replay attack detection, end-to-end detection, deep convolutional neural networks, attention mechanism, biometric authentication security
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.