摘要
网络空间是一个开放的环境,允许用户自由地进行通信、资源共享和在线活动,这种开放性容易导致网络被黑客攻击。通过识别网络空间安全实体,可以分析其行为模式和特征,从而预测和预防未来的网络攻击。因此,研究融合广义同余神经网络(GCNN)与门挖循环单元(GRU)的网络空间安全实体识别方法。建立变换器的双向编码器表征量(BERT)预训练语言模型,提取文本句子中的字嵌入、段嵌入以及位置嵌入表示,获得语料库的特征信息以及实体的具体位置。由双向GRU获取文本句子的上下文语义信息,经过GCNN的卷积运算后,输出最佳特征向量,即网络空间安全实体识别结果。通过实验结果表明,所提方法的准确率和召回率始终在95%以上,F_(1)值最接近1,网络空间安全实体识别结果精准,可有效保护网络安全。
The cyberspace is an open environment,it allows users to freely communicate,share resources,and engage in online activities.This openness makes the network vulnerable to attacks by hackers.By identifying cyberspace security entities,their behavior patterns and characteristics can be analyzed to predict and prevent future cyber attacks.To this end,a cyberspace security entity recognition method integrating generalized congruence neural network(GCNN)and gated recurnent unit(GRU)is studied.A bidirectional encoder representations from transformers(BERT)pretrained language model is established to extract word embeddings,segment embeddings and positional embeddings from text sentences,and obtain feature information from the corpus as well as the specific positions of entities.The context semantic information of the text sentence is obtained by bidirectional GRU,and after the convolution operation by GCNN,the optimal feature vector is output,which is the result of cyberspace security entity recognition.The experimental results show that the accuracy and recall of the proposed method are always above 95%,and the F_(1) value is closest to 1.The cyberspace security entity recognition results are accurate,and can effectively protect network security.
作者
张国章
ZHANG Guozhang(Guangdong Construction Engineering Group Holding Co.,Ltd.,Guangzhou 510000,China)
出处
《微型电脑应用》
2024年第11期242-246,共5页
Microcomputer Applications
