Adversarial Attack

How to tackle adversarial attack in rasa for intent classification ?

I know there are some methods which related to augment more data.

But can I prevent adversarial attack by rasa 's DIET ?

What is current research direction in rasa ?

Some survey I have read: here and here