Audio Adversial Attack

FGSM, I-FGSM, and PGD algorithms are implemented to attack ASR models.

Getting to know:

The ASR_Experiment notebook contains the implemenation of FGSM, I-FGSM and PGD to attack.

Auto_encoder notebook attacks the Deepspeech model with variouse audio adversials that were generated earlier. It also contains contiguency defence mechanisum with the help of densosing auto encoders.