The aim of the project is to design ML models which are robust against adversarial attacks.  The specific use-case is to make  credit scoring models robust. We will not only explore model robustness against adversarial attacks but also concept-drift and covariate-shift.