基于特征学习的无线传感网络入侵行为检测方法

传统的机器学习算法在无线传感网络入侵行为检测中应用效果不理想,Recall(召全率)和F1-score(召全率与精准率的调和平均数)较低,针对现行方法存在的不足和缺陷,提出基于特征学习的无线传感网络入侵行为检测方法。利用时间戳马尔可夫模型对无线传感网络进行分段,实现对网络入侵数据局部特征编码,利用深度学习网络,学习网络入侵行为特征,对网络数据标签进行数值化和归一化处理,根据入侵特征对网络行为进行分类,识别检测入侵行为,以此实现基于特征学习的无线传感网络入侵行为检测。经实验证明,设计方法Recal在95%以上,F1-score在90%以上,检测精度较高,在无线传感网络入侵行为检测方面具有良好的应用前景。

The application effect of traditional machine learning algorithms in intrusion detection of wireless sensor networks is not ideal,with low Recall(recall)and F1 score(harmonic average of recall and accuracy).In response to the shortcomings and shortcomings of current methods,a feature learning based intrusion detection method for wireless sensor networks is proposed.Using timestamp Markov model of wireless sensor network,realize the local characteristics of network intrusion data coding,using deep learning network,learning network intrusion behavior characteristics,numerical and normalized processing,according to the intrusion characteristics of network behavior,identification and detection behavior,to realize the wireless sensor network intrusion behavior detection based on feature learning.The experiment proved that the design method Recall is above 95%and F1-score is above 90%,with high detection accuracy and has good application prospect in detecting intrusion behavior of wireless sensor network.