Exploring Adversarial Attacks on Neural Networks: An Explainable Approach