Spoofing detection systems are typically trained using diverse recordings from multiple speakers, often assuming that the resulting embeddings are independent of speaker identity. However, this assumption remains unverified. In this paper, we investigate the impact of speaker information on spoofing detection systems. We propose two approaches within our Speaker-Invariant Multi-Task framework, one that models speaker identity within the embeddings and another that removes it. SInMT integrates multi-task learning for joint speaker recognition and spoofing detection, incorporating a gradient reversal layer. Evaluated using four datasets, our speaker-invariant model reduces the average equal error rate by 17% compared to the baseline, with up to 48% reduction for the most challenging attacks (e.g., A11).
Figures & tables
Figure 1: Our model architecture with dual classifiers and GRL.
EER
Model
ITW
ASV5 eval
ASV21LA
ASV21DF
Pooled
AASIST
7.03
5.54
13.66
9.60
8.95
Conformer
5.69
3.85
12.49
10.40
8.10
MHFA
4.31
4.64
12.14
8.58
7.41
MHFA-spk
3.76
5.29
8.67
8.41
6.53
MHFA-IVspk
3.58
4.98
8.41
7.57
6.13
Table 1: Model performance comparison in term of Equal Error Rate (EER).
Model
A07
A08
A09
A10
A11
A12
A13
A14
A15
A16
A17
A18
A19
Pooled
MHFA
1.54
1.91
0.76
20.77
17.02
3.45
4.75
3.82
1.49
4.32
7.01
19.67
37.57
12.14
MHFA-spk
1.07
1.14
0.67
14.40
10.30
3.72
6.61
1.94
1.15
3.72
4.93
11.18
28.61
8.67
MHFA-IVspk
0.88
1.15
0.45
12.56
8.76
3.52
3.61
2.36
1.12
3.32
4.76
15.45
24.85
8.41
Table 2: Model performance (% EER) of different models, evaluated on ASVspoof 2021 LA break down: attack types (A07-19). The pooled indicates the average performance across all categories.
Figure 2: t-SNE visualization of embeddings from MHFA, MHFA-spk and MHFA-IVspk models, showcasing audio representations of ten distinct speakers from the ASVspoof 5 dataset, each represented by a unique color.
Sophisticated generative speech technology can undermined the reliability of voice biometrics. While spoofing detection systems excel when assessed under in-domain conditions, generalisation to out-of-domain settings is often poor. In this paper, we show that such issues could be caused by speaker bias, where models learn individual voice traits rather than markers of manipulation or generation. We propose a teacher-student framework for speaker-invariant spoofing detection that disentangles identity without requiring speaker labels. We leverage a pre-trained speaker recognition teacher to guide a student model via a gradient reversal layer. To control the balance between suppressing cues related to voice identity with the preservation of those related to spoofing detection, we integrate a Variational Information Bottleneck. Evaluations across nine datasets show our model achieves a 25.7% relative reduction to the EER compared to the MHFA baseline.
Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier +1
Laboratoire Informatique d’Avignon, Avignon Universite, France · EURECOM, Sophia Antipolis, France
The performance of speech spoofing detection often varies across different training and evaluation corpora. Leveraging multiple corpora typically enhances robustness and performance in fields like speaker recognition and speech recognition. However, our spoofing detection experiments show that multi-corpus training does not consistently improve performance and may even degrade it. We hypothesize that dataset-specific biases impair generalization, leading to performance instability. To address this, we propose an Invariant Domain Feature Extraction (IDFE) framework, employing multi-task learning and a gradient reversal layer to minimize corpus-specific information in learned embeddings. The IDFE framework reduces the average equal error rate by 20% compared to the baseline, assessed across four varied datasets.
Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier +1
Laboratoire d’informatique d’Avignon, France · EURECOM, Sophia Antipolis, France
Rapid advancements in generative speech technology have compromised the reliability of voice biometrics. While current spoofing detectors excel when assessed under in-domain conditions, generalisation to out-of-domain settings is often poor. We show that this can be due to linguistic bias. A reliance on linguistic cues observed in training data can then compromise robustness to cross-data. We propose a linguistic-invariant spoofing detection framework utilizing teacher-student adversarial learning. The linguistic-aware teacher model, pre-trained on linguistic content of an external dataset, guides the student detector via gradient reversal to minimize the linguistic information. To prevent the inadvertent removal of non-linguistic cues, we incorporate a Variational Information Bottleneck to enable suppression of principal cues. Across nine DF Arena datasets, our method achieves up to a 36.2% relative reduction in the EER compare to the baseline.
Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier +1
Laboratoire Informatique d’Avignon, Avignon Universite, France · EURECOM, Sophia Antipolis, France