基于深度学习的无线电信号对抗样本检测研究

Research on detection of radio signal adversarial samples based on deep learning

针对无线电信号的攻击愈来愈频繁的情况,本文在数据流形理论基础上,使用深度神经网络(DNN)检测无线电信号对抗样本及其攻击方法。首先使用5种不同攻击方法对无线电信号进行攻击产生对抗样本,其次使用3种不同的神经网络检测对抗样本,最后用残差神经网络(ResNet)检测对抗样本的攻击方法。在信噪比(SNR)为30 d B和20 dB的无线电信号数据上的实验结果表明,本文所使用的残差神经网络检测精度接近100%,在信噪比为10 dB的无线电信号数据上的检测精度仍然在90%以上。结果表明本文所用的残差神经网络能有效检测无线电信号的对抗样本及其攻击方法。

Aiming at the problem of the increasing frequent attacks on radio signals,based on data manifold theory,the deep neural network(DNN)is used to detect radio signal adversarial samples and their attack methods.First,five different attack methods are used to attack radio signals to generate adversarial samples.Second,three different neural networks are used to detect adversarial samples.Last,the residual neural network(ResNet)is used to detect adversarial samples’attack methods.The experimental results on radio signal data with signal-to-noise-ratio(SNR)of 30 dB and 20 dB show that the detection accuracy of the residual neural network used in this paper is close to 100%,while the experimental results on radio signal data with SNR of 10 dB show that the detection accuracy is still above 90%.The results show that the residual neural network used in this paper can effectively detect the adversarial samples of radio signals and their attack methods.