基于生成对抗网络的UWSNs恶意节点检测方法

Malicious Nodes Detection Method for UWSNs Based on Generative Adversarial Network

水下无线传感器网络(Underwater Wireless Sensor Networks,UWSNs)节点通常部署于无人值守的恶劣环境中,恶意节点的存在对网络的安全构成了极大威胁。为了在缺乏节点行为证据的情况下有效检测恶意节点,提出了一种基于生成对抗网络(Generative Adversarial Network,GAN)的恶意节点检测方法。首先,构建了基于GAN的异常预测模型;其次,收集数据、链路、能量等多维信任证据,构建出训练数据集;最后,训练异常预测模型,并通过信任决策检测恶意节点。仿真结果表明,该方法可以在不需要大量信任证据的情况下有效地进行恶意节点检测,并且与同类型方法相比,所提方法具有较高的恶意节点检测率、通信成功率及较低的假阳率。

UWSNs(Underwater Wireless Sensor Networks)nodes are usually deployed in unattended or harsh environments,and the presence of malicious nodes poses a great threat to the security of the networks.To effectively detect malicious nodes in the absence of evidence about node behavior,a malicious node detection method based on GAN(Generative Adversarial Network)is proposed.First,a GAN-based anomaly prediction model is constructed.Then,multi-dimensional trust evidences regarding data,link,and energy are collected to construct training dataset.Finally,the anomaly prediction model is trained and malicious nodes are detected through trust decision.Simulation results indicate that the method can effectively perform malicious node detection without large numbers of trust evidences.In addition,the proposed method has a better performance on malicious node detection rate,communication success rate,and low false positive rate than other state-of-the-art methods.