Hear "No Evil", See "Kenansville"*: Efficient and Transferable Black-Box Attacks on Speech Recognition and Voice Identification Systems.
Hadi AbdullahMuhammad Sajidur RahmanWashington GarciaKevin WarrenAnurag Swarnim YadavTom ShrimptonPatrick TraynorPublished in: IEEE Symposium on Security and Privacy (2021)
Keyphrases
- speech recognition
- black box
- black boxes
- speech understanding
- speech recognition systems
- speech synthesis
- speech recognizers
- language model
- hidden markov models
- pattern recognition
- speech recognition errors
- automatic speech recognition
- voice activity detection
- white box
- noisy environments
- speech processing
- speech recognizer
- neural network
- white box testing
- feature selection
- speech signal
- integration testing
- gaussian mixture model
- signal processing
- speech recognition technology
- high level
- multimedia