Disappeared Command: Spoofing Attack On Automatic Speech Recognition Systems with Sound Masking
This addresses a security problem for users of voice interfaces and smart devices, but it appears incremental as it builds on known vulnerabilities in DNN-based systems.
The paper tackles the vulnerability of automatic speech recognition (ASR) systems to spoofing attacks using sound masking, highlighting that deep learning-based ASR can be easily disturbed by slight disturbances, leading to false recognition and posing dangers for voice-controlled applications.
The development of deep learning technology has greatly promoted the performance improvement of automatic speech recognition (ASR) technology, which has demonstrated an ability comparable to human hearing in many tasks. Voice interfaces are becoming more and more widely used as input for many applications and smart devices. However, existing research has shown that DNN is easily disturbed by slight disturbances and makes false recognition, which is extremely dangerous for intelligent voice applications controlled by voice.