Protecting the Neural Networks against FGSM Attack Using Machine Unlearning