Enhancing Deep Learning Security Through Explainability And Robustness