Analysing Replay Spoofing Countermeasure Performance Under Varied Conditions

Chettri, Bhusan; Sturm, Bob L.; Benetos, Emmanouil

doi:10.1109/mlsp.2018.8516968

Cited by 9 publications

(6 citation statements)

References 14 publications

(31 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We train two SVMs using i-vectors (model I) and the longterm-average-spectrum (LTAS) feature (model J) since they have shown good performance on prior spoofing datasets [30,31,32]. Inspired from [12] we fuse multiple i-vectors in our approach, each based on complimentary hand-engineered features, and manage to improve performance over a single i-vector based SVM.…”

Section: Svmmentioning

confidence: 99%

Ensemble Models for Spoofing Detection in Automatic Speaker Verification

et al. 2019

Self Cite

View full text Add to dashboard Cite

Spectrograms -time-frequency representations of audio signals -have found widespread use in neural network-based spoofing detection. While deep models are trained on the fullband spectrum of the signal, we argue that not all frequency bands are useful for these tasks. In this paper, we systematically investigate the impact of different subbands and their importance on replay spoofing detection on two benchmark datasets: ASVspoof 2017 v2.0 and ASVspoof 2019 PA. We propose a joint subband modelling framework that employs n different sub-networks to learn subband specific features. These are later combined and passed to a classifier and the whole network weights are updated during training. Our findings on the ASVspoof 2017 dataset suggest that the most discriminative information appears to be in the first and the last 1 kHz frequency bands, and the joint model trained on these two subbands shows the best performance outperforming the baselines by a large margin. However, these findings do not generalise on the ASVspoof 2019 PA dataset. This suggests that the datasets available for training these models do not reflect real world replay conditions suggesting a need for careful design of datasets for training replay spoofing countermeasures.

show abstract

Section: Svmmentioning

confidence: 99%

Ensemble Models for Spoofing Detection in Automatic Speaker Verification

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…Mishra et al [5] proposed the first adaption of LIME for MIR, termed SoundLIME (SLIME), which segments the spectrogram into time, frequency, or time-frequency segments. SLIME was demonstrated on the task of singing voice detection [18] and used for analysing a replay spoofing detection system [6]. Haunschmid et al used other types of interpretable features (super pixels [7], source sep-aration estimates [8,9]) for explaining the predictions of a variety of models, including music taggers [8,9] and a content-based music recommender system [10].…”

Section: Local Interpretable Model-agnostic Explanationsmentioning

confidence: 99%

“…A plethora of explanation methods ("explainers") have been originally developed for text or image data and adapted to the audio domain [1,2], or specifically introduced for MIR systems [3]. Most notably, different versions of Local Interpretable Model-agnostic Explanations (LIME), a posthoc explainer [4], have been used to explain models in a variety of MIR tasks [5][6][7][8][9][10][11].…”

Section: Introductionmentioning

confidence: 99%

On the Veracity of Local, Model-agnostic Explanations in Audio Classification: Targeted Investigations with Adversarial Examples

Praher¹,

Prinz²,

Flexer³

et al. 2021

Preprint

View full text Add to dashboard Cite

Local explanation methods such as LIME have become popular in MIR as tools for generating post-hoc, model-agnostic explanations of a model's classification decisions. The basic idea is to identify a small set of humanunderstandable features of the classified example that are most influential on the classifier's prediction. These are then presented as an explanation. Evaluation of such explanations in publications often resorts to accepting what matches the expectation of a human without actually being able to verify if what the explanation shows is what really caused the model's prediction. This paper reports on targeted investigations where we try to get more insight into the actual veracity of LIME's explanations in an audio classification task. We deliberately design adversarial examples for the classifier, in a way that gives us knowledge about which parts of the input are potentially responsible for the model's (wrong) prediction. Asking LIME to explain the predictions for these adversaries permits us to study whether local explanations do indeed detect these regions of interest. We also look at whether LIME is more successful in finding perturbations that are more prominent and easily noticeable for a human. Our results suggest that LIME does not necessarily manage to identify the most relevant input features and hence it remains unclear whether explanations are useful or even misleading. * Equal contribution.

show abstract

“…rectangular subparts of spectrograms) as an explainer, as this is the most prominent method in audio (cf. [19][20][21][22]). Then, in Sect.…”

Section: Introductionmentioning

confidence: 99%

Constructing adversarial examples to investigate the plausibility of explanations in deep audio and image classifiers

Hoedt¹,

Praher²,

Flexer³

et al. 2022

Neural Comput & Applic

View full text Add to dashboard Cite

Given the rise of deep learning and its inherent black-box nature, the desire to interpret these systems and explain their behaviour became increasingly more prominent. The main idea of so-called explainers is to identify which features of particular samples have the most influence on a classifier’s prediction, and present them as explanations. Evaluating explainers, however, is difficult, due to reasons such as a lack of ground truth. In this work, we construct adversarial examples to check the plausibility of explanations, perturbing input deliberately to change a classifier’s prediction. This allows us to investigate whether explainers are able to detect these perturbed regions as the parts of an input that strongly influence a particular classification. Our results from the audio and image domain suggest that the investigated explainers often fail to identify the input regions most relevant for a prediction; hence, it remains questionable whether explanations are useful or potentially misleading.

show abstract

Analysing Replay Spoofing Countermeasure Performance Under Varied Conditions

Cited by 9 publications

References 14 publications

Ensemble Models for Spoofing Detection in Automatic Speaker Verification

Ensemble Models for Spoofing Detection in Automatic Speaker Verification

On the Veracity of Local, Model-agnostic Explanations in Audio Classification: Targeted Investigations with Adversarial Examples

Constructing adversarial examples to investigate the plausibility of explanations in deep audio and image classifiers

Contact Info

Product

Resources

About