Adversarial attacks against deep learning models have gained significant\nattention and recent works have proposed explanations for the existence of\nadversarial examples and techniques to defend the models against these attacks.\nAttention in computer vision has been used to incorporate focused learning of\nimportant features and has led to improved accuracy. Recently, models with\nattention mechanisms have been proposed to enhance adversarial robustness.\nFollowing this context, this work aims at a general understanding of the impact\nof attention on adversarial robustness. This work presents a comparative study\nof adversarial robustness of non-attention and attention based image\nclassification models trained on CIFAR-10, CIFAR-100 and Fashion MNIST datasets\nunder the popular white box and black box attacks. The experimental results\nshow that the robustness of attention based models may be dependent on the\ndatasets used i.e. the number of classes involved in the classification. In\ncontrast to the datasets with less number of classes, attention based models\nare observed to show better robustness towards classification.\n