Research On Construction Of Knowledge Graph Of Cyber Security Events Based On Deep Learning

With the continuous development of the new generation of network information industry,cyberspace security threats are becoming increasingly serious.Traditional network security data mining methods have been unable to meet future development requirements.With the development and information of key technologies such as machine learning,deep learning,and natural language processing,With explosive growth,how to mine useful information from massive amounts of cyber security incident data becomes a problem.Constructing an event knowledge map can help decision makers understand the attack chain and future trends of cyber security incidents,and become an important research direction for the development of event knowledge map.Therefore,this article takes cyber security incidents as the research object,and designs and implements a knowledge graph system based on deep learning.The cyber security incidents that this article focuses on are events that cause harm to the network and information systems or the data in them,and negatively affect the society,such as network unwanted program events,network attack events,and network information destruction events.Based on the in-depth summary and research of the existing results,a method for constructing a knowledge map of cyber security incidents is proposed.The main research contents are as follows:(1)In terms of data collection and storage,this article designs and implements a crawler architecture.By combining strategies and technical methods such as web crawlers,lxml,regular expressions,requests,and multi-threading concurrency,it solves problems such as slow data collection and achieves Efficient and automatic acquisition of corpora.(2)In terms of knowledge extraction,based on the study of deep learning algorithms for knowledge extraction,this paper focuses on the feature selection models of Word2 Vec,GloVe,FastText word vectors,and proposes a knowledge extraction analysis combining word vector feature selection models with deep learning.model.In this paper,the Bilstm-crf algorithm is used for comparison experiments,and the optimal knowledge extraction model is obtained through model tuning and repeated experiments.Experimental results show that the model proposed in this paper has an accuracy of 84% and an F value of 0.81.The model has a good recognition effect on the type of event information entities in the corpus.(3)In terms of knowledge graph construction,a framework and design scheme for the construction of a knowledge graph for cyber security incidents is established.Use Neo4 j graph data for knowledge storage and query,and implement data collection,data storage,knowledge extraction,and knowledge graph display for each functional module.The knowledge map of network security events has been completed.