Whisper Smarter, not Harder: Adversarial Attack on Partial Suppression
This work addresses security vulnerabilities in ASR systems, which are widely deployed, but it is incremental as it builds on existing attack methods.
The paper tackles the problem of adversarial attacks on Automatic Speech Recognition (ASR) models by investigating how to increase their imperceptibility, finding that relaxing the optimization objective from complete to partial suppression reduces perceptibility, and it shows a low-pass filter can serve as an effective defense.
Currently, Automatic Speech Recognition (ASR) models are deployed in an extensive range of applications. However, recent studies have demonstrated the possibility of adversarial attack on these models which could potentially suppress or disrupt model output. We investigate and verify the robustness of these attacks and explore if it is possible to increase their imperceptibility. We additionally find that by relaxing the optimisation objective from complete suppression to partial suppression, we can further decrease the imperceptibility of the attack. We also explore possible defences against these attacks and show a low-pass filter defence could potentially serve as an effective defence.