Skip to main content

Table 3 SNR(dB) of adversarial examples on commercial cloud speech-to-text APIs

From: Towards the transferable audio adversarial attack via ensemble methods

API service

Audio length

DS

ITRA

CS

DW

Occam

DC

RGE

DGWE

Aliyun-API

4s

26.47

–

19.02

17.33

17.84

–

13.51

18.01

Xfyun-API

11.22

14.20

12.85

Baiduyun-API

–

12.21

13.34

Tencent-API

16.70

–

–

Microsoft-API

10.74

–

–

Average

4s

26.47

–

19.02

17.33

14.13

 

13.31

14.73