add two papers

[EdoardoMosca]

“That Is a Suspicious Reaction!”: Interpreting Logits Variation to Detect NLP Adversarial Attacks
Detecting Word-Level Adversarial Text Attacks via SHapley Additive exPlanations

[yangalan123]

Hi @EdoardoMosca , thanks for your contribution! I have added your two papers in the defense papers section as I think detecting adversarial attacks is more relevant on the defense side. I have also attached your github repo link and pdf link for readers' reference. Let me know if you have further questions.