Countermeasures Against Adversarial Examples in Radio Signal Classification

Deep learning algorithms have been shown to be powerful in many communication network design problems, including that automatic modulation classification. However, they are vulnerable carefully crafted attacks called adversarial examples. Hence, the reliance of wireless networks on deep poses a serious threat security and operation networks. In this letter, we propose for first time countermeasure against examples Our is based neural rejection technique, augmented by label smoothing Gaussian noise injection, allows detect reject with high accuracy. results demonstrate proposed can protect deep-learning classification systems