鼠标行为HHT变换的工业互联网用户身份认证

User authentication of industrial internet based on HHT transform of mouse behavior

工业互联网的快速发展引发了对网络安全的广泛关注,终端用户身份认证技术成为研究热点。根据工业互联网人机交互特点,设计了实验网站,收集了该网站24名用户两年半的非受控环境下鼠标行为数据作实例,采用希尔伯特黄变换(HHT,Hilbert-Huang transform)提取鼠标行为信号频域特征,结合时域特征,形成163维时频域联合特征矩阵,用于表征用户鼠标行为模式特征。使用Bagged tree、支持向量机(SVM,support vector machine)、Boost tree和K最邻近(KNN,K-nearest neighbor)算法构建网络用户身份认证模型,对比数据测试结果表明,Bagged tree算法在本案例中内部检测效果最佳,平均错误接受率(FAR,false acceptance rate)为0.12%、平均错误拒绝率(FRR,false rejection rate)为0.28%;外部检测中,平均FAR为1.47%。相较于传统鼠标动力学方法,使用HHT提取鼠标行为频域信息能更好地实现终端用户身份认证,为保障工业互联网安全提供有效的技术支撑。

The rapid development of the industrial internet had caused widespread concern about the network security,and the end-user authentication technology was considered a research hotspot.According to the characteristics of human-computer interaction in industrial internet,an experimental website was designed.24 users'mouse behavior data in an uncontrolled environment were collected within 2.5 years to conduct case studies.Hilbert-Huang transform(HHT)was used to extract frequency domain features of mouse behavior signals,combined with time domain features to form a time-frequency joint domain feature matrix of 163-dimensional to characterize user mouse behavior patterns.Bagged tree,support vector machine(SVM),Boost tree and K-nearest neighbor(KNN)were used to build a user authentication model,and the comparison result showed that the Bagged tree had the best internal detection effect in this case,with an average false acceptance rate(FAR)of 0.12%and an average false rejection rate(FRR)of 0.28%.In external detection,the FAR was 1.47%.Compared with the traditional mouse dynamics method,the frequency domain information of mouse behavior extracted by HHT can better realize the user authentication,and provide technical support the security of the industrial internet.