| 摘要: |
| 利用深度学习解决无线通信识别问题的研究越来越多。为了建立安全可靠的深度神经网络模型，有必要研究其对抗攻击技术。在介绍对抗样本概念和攻击算法的基础上，提出了无线通信智能识别神经网络的对抗攻击模型；对近年来无线通信中智能识别神经网络的对抗攻击方法进行了对比分析，给出了当前对抗攻击方法的局限性和后续研究重点；最后展望了未来的研究方向。 |
| 关键词: 无线通信 智能识别 神经网络 对抗攻击 对抗样本 |
| DOI：10.20079/j.issn.1001-893x.220221002 |
|
| 基金项目: |
|
| Survey on adversarial attack of intelligent recognition neural network in wireless communications |
| HAN Chao,QIN Ruoxi,WANG Linyuan,CUI Weijia,YAN Bin |
| (Institute of Information System Engineering,PLA Strategic Support Force Information Engineering University,Zhengzhou 450001,China) |
| Abstract: |
| With the increasing research on wireless communication identification based on deep learning,it is necessary to study its adversarial attack technology to establish a safe and reliable deep neural network model.The concept of adversarial samples and related attack algorithms are introduced,and then an adversarial attack model of intelligent recognition neural network in wireless communications is proposed.Through the comparative analysis of the adversarial attack methods of the wireless communication intelligent identification model in recent years,the limitations of the current adversarial attack methods and the follow-up research focuses are given.Finally,the future research direction is prospected. |
| Key words: wireless communications intelligent recognition neural network adversarial attack adversarial sample |
