Towards logical specification of adversarial examples in machine learning | IEEE Conference Publication | IEEE Xplore