Adversarial Attack and Defense for Commercial
Black-box Multilingual Speech Recognition Systems
Embedded Files
1. AEs for Survey
1. AEs for Survey
multi_Survey_AEs.zip2. AEs for the Digital World Attack
2. AEs for the Digital World Attack
2.1. Digital world attack (attack Aliyun API)
muti_AEsofaliyunAPI.mp42.2. Digital world attack (attack Tencentyun API)
muti_AEsfortencentyun.mp43. AEs for Physical world attack
3. AEs for Physical world attack
3.1. Physical world attack demos (indoor)
3.1. Physical world attack demos (indoor)
(1) Siri recognizes the sound as “播放音乐”， which means “play music” in English.
Siri_AE_01180001-播放音乐-9d10.mp4(2) Siri recognizes the sound as “打开QQ”， which means “open QQ” in English.
Siri_AE_011800006-打开QQ_10d10.mp4(3) Siri recognizes the sound as “打开QQ音乐”， which means “open QQ music” in English.
Siri_AE_01180014-打开QQ音乐-5d10.mp4(4) Siri recognizes the sound as “我要听歌”， which means “ I want to listen to songs” in English.
Siri_AE_01180007-我要听歌_10d10.mp4(5) Siri recognizes the sound as “支付宝付款”， which means “AliPay payment” in English.
Siri_AE_01180002-支付宝付款-8d10.mp4(6) Siri recognizes the sound as “安装抖音”， which means “install Tiktok ” in English.
Siri_AE_01100016-安装抖音-10d10.mp4(7) Xiaoaitongxue recognizes the sound as “今天天气怎么样”， which means “what is the weather today” in English.
xiaoai_0190020-今天天气怎么样-6d10.mp4(8) Xiaoaitongxue recognizes the sound as “把电视机打开”， which means “turn on the TV” in English.
xiaoai_01190006-把电视机打开-8d10.mp4(9) Xiaoaitongxue recognizes the sound as “打开浏览器”， which means “open browser” in English.
xiaoai_01190003-打开浏览器-6d10.mp4(10) Xiaoaitongxue recognizes the sound as “直接关机”， which means “directly close the phone” in English.
xiaoai_01190017-直接关机-4d10.mp4(11) Xiaoaitongxue recognizes the sound as “打开车灯”， which means “turn on the lights of the car ” in English.
xiaoai_01190014-打开车灯-4d10.mp4(12) Xiaoaitongxue recognizes the sound as “金融诈骗”， which means “financial fraud” in English.
xiaoai_01190023-金融诈骗-4d10.mp43.2. Physical world attack demos (outdoor)
3.2. Physical world attack demos (outdoor)
(1) Siri recognizes the sound as “截屏”， which means “Screen cut” in English.
01_截屏.mp4(2) Siri recognizes the sound as “开始录音”， which means “start record” in English.
02-开始录音.mp4(3) Siri recognizes the sound as “打开电视”, which means “turn on the TV”.
04-打开电视.mp4(4)Siri recognizes the sound as “安装快手”, which means “install kwaishou”.
03-安装快手.mp4(5) Siri recognizes the sound as “打电话给小陈”, which means “call Xiao Chen”.
05-打电话给小陈.mp44. AEs for Others
4. AEs for Others
4.1. AEs of wav2vec2 transfer to Aliyun API service
4.1. AEs of wav2vec2 transfer to Aliyun API service
01_AEs_wav2vec2_transfer_Aliyun.mp44.2. AEs of DataTang transfer to Aliyun API service
4.2. AEs of DataTang transfer to Aliyun API service
04_AEs_DataTang_transfer_Aliyun.mp44.3. AEs of DataTang transfer to Azure and Aliyun API service
4.3. AEs of DataTang transfer to Azure and Aliyun API service
02_AEs_DataTang_transfer_Aliyun_Azure_demo.mp44.4. AEs of DataTang transfer to Iflytec and Aliyun API service
4.4. AEs of DataTang transfer to Iflytec and Aliyun API service
03_AEs_DataTang_transfer_Aliyun_Iflytec(xunfei)_demo.mp4Page updated
Google Sites
Report abuse