2020
DOI: 10.2478/popets-2021-0012
|View full text |Cite
|
Sign up to set email alerts
|

The Audio Auditor: User-Level Membership Inference in Internet of Things Voice Services

Abstract: With the rapid development of deep learning techniques, the popularity of voice services implemented on various Internet of Things (IoT) devices is ever increasing. In this paper, we examine user-level membership inference in the problem space of voice services, by designing an audio auditor to verify whether a specific user had unwillingly contributed audio used to train an automatic speech recognition (ASR) model under strict black-box access. With user representation of the input audio data and their corres… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
4
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
3
3
1

Relationship

0
7

Authors

Journals

citations
Cited by 14 publications
(4 citation statements)
references
References 34 publications
0
4
0
Order By: Relevance
“…MIA on other models. Other MIA targeted conventional classification [10], [15], [11], [16], [12], [17], [13], speech recognition [18], [19], generative [21], [22], regression models [14], etc. These MIA cannot be ported to speaker recognition due to the distinct training paradigm and architecture of speaker recognition.…”
Section: Membership Inference Attackmentioning
confidence: 99%
See 1 more Smart Citation
“…MIA on other models. Other MIA targeted conventional classification [10], [15], [11], [16], [12], [17], [13], speech recognition [18], [19], generative [21], [22], regression models [14], etc. These MIA cannot be ported to speaker recognition due to the distinct training paradigm and architecture of speaker recognition.…”
Section: Membership Inference Attackmentioning
confidence: 99%
“…While numerous studies have confirmed the feasibility of MIA in various applications of DNNs, including image classification [10], [14], [15], [11], [16], [12], [17], speech recognition [18], [19], language models [20], and generative models [21], [22], MIA against SRSs has not been considered yet. Considering the wide spread of voices across social media platforms, online meetings, and voice-enabled smart devices, users' voice data may be collected and used for training SRSs without their consent.…”
Section: Introductionmentioning
confidence: 99%
“…In the literature, there are only a few studies on user-level MI attacks. In Miao et al (2021), the authors investigate MI attacks on speech recognition task to infer if any users' data (voice samples) have been used during training. In Song & Shmatikov (2019), the authors propose a user-level MI attack on text generative models.…”
Section: Membership Inferencementioning
confidence: 99%
“…Baselines: To the best of our knowledge, there is no user-level MI attack on metric embedding learning. The two user-level MI attacks in literature (Song & Shmatikov, 2019;Miao et al, 2021) require generative models where the victim model's output is a word. Hence, there is no trivial way to adopt them for metric embedding scenario.…”
Section: Experimental Settingsmentioning
confidence: 99%