word2vec基础上的配电网恶意控制指令检测

Detection of Malicious Control Instruction in Distribution Network Based on word2vec

提出了一种根据上下文数据关系建立的word2vec算法。针对大量访问数据来建立白名单模型,通过对配电网上下文测量信息和控制信息的挖掘和数据驱动实现恶意控制指令的快速检测,获得白名单模型中的不符合项作为异常。利用孤立森林算法建立上下文关系的孤立树,从而实现对各测试样本的分类和训练,采用CBOW神经网络模型将中心词汇后验概率作为输出层,获得不同样本集下的监测精确度和准确率。最后在建立的配电网仿真平台上对word2vec进行了数据挖掘和计算,验证了算法具有高准确率和低误警率。

In the paper,a word2vec algorithm based on context data relation is proposed.The whitelist model is established for a large number of accessing data.By mining and data driving the context measurement information and control information of distribution network,the malicious control instructions can be quickly detected,and the nonconformance items in the whitelist model can be obtained as exceptions.The isolated tree of context relation is established by using the isolated forest algorithm to realize the classification and training of each test sample.The CBOW neural network model is used to classify the central vocabulary.A posteriori probability is used as an output layer to obtain the monitoring accuracy under different sample sets.Finally,the data mining and calculation of word2vcc are carried out on the simulation platform of distribution network,which proves that the algorithm has high accuracy and low false alarm rate.