Universal adversarial examples in speech command classification

Jon Vadillo, Roberto Santana

Adversarial examples are inputs intentionally perturbed with the aim of forcing a machine learning model to produce a wrong prediction, while the changes are not easily detectable…