摘要
考虑到基于梯度的对抗攻击生成算法在实际通信系统部署中面临着因果性问题,提出了一种因果性对抗攻击生成算法。利用长短期记忆网络的序列输入输出特征与时序记忆能力,在满足实际应用中存在的因果性约束前提下,有效提取通信信号的时序相关性,增强针对无人通信系统的对抗攻击性能。仿真结果表明,所提算法在同等条件下的攻击性能优于泛用对抗扰动等现有的因果性对抗攻击生成算法。
A causality adversarial attack generation algorithm was proposed in response to the causality issue of gradient-based adversarial attack generation algorithms in practical communication system.The sequential input-output features and temporal memory capability of long short-term memory networks were utilized to extract the temporal correlation of communication signals while satisfying practical causality constraints,and enhance the adversarial attack performance against unmanned communication systems.Simulation results demonstrate that the proposed algorithm outperforms existing causality adversarial attack algorithms,such as universal adversarial perturbation,under identical conditions.
作者
禹树文
许威
姚嘉铖
YU Shuwen;XU Wei;YAO Jiacheng(National Mobile Communications Research Laboratory,Southeast University,Nanjing 210096,China;Purple Mountain Laboratories,Nanjing 211111,China)
出处
《通信学报》
EI
CSCD
北大核心
2024年第1期54-62,共9页
Journal on Communications
基金
国家自然科学基金资助项目(No.62022026,No.62211530108)
中央高校基本科研业务费专项资金资助项目(No.2242022K60002,No.2242023K5003)。
关键词
智能通信系统
对抗攻击
深度学习
因果系统
长短期记忆网络
intelligent communication system
adversarial attack
deep learning
causal system
long short-term memory network