Research On DDoS Attack Prediction Model Based On Hacker Forum Topics

While the development of information technology has facilitated efficient online socializing,online shopping and online payments,it has also provided the tools and means for hackers to attack.Among them,hackers are constantly launching DDoS(Distributed Denial of Service)attacks using malicious tools.Hacker forums,as one of the social platforms for hackers to gather and communicate,hide a lot of malicious attack resources and dark acts.By posting threads to communicate with others and share their experiences,hackers also spread free DDoS attack tools and other attack resources while socializing,resulting in unpredictable DDoS attacks.Research has focused on the user behavior of hackers and the network of member interactions in forums,and less on the hacking tools hidden in the text of posts,and Only some scholars have made the connection to actual cyber attacks.Therefore,it is important to study actual cyber attacks from the platform of hacking forums.Based on the above issues,this thesis takes the hacker forum as the data source for the study from the perspective of DDoS attacks,and crawls the post texts in the hacker forum.Faced with a large number of mixed short text sets,this thesis takes text topic extraction as the main task and uses the LDA(Latent Dirichlet Allocation)topic model as the base model to establish a hacker forum-based DDoS attack tool topic identification and its evolution model to identify the hackers' existing DDoS attack tools.At the same time,in order to establish a connection between hackers' online discussions and actual attacks,and to explore the deeper semantics in the discussions,this thesis develops a DDoS attack prediction model that blends the two from the perspective of hacker tools,from both the thematic and sentiment perspectives,and compares it with other prediction models.In the study of DDoS attack tool theme identification,10 topics are co-categorized in this thesis,and the domain boundary between each topic is clear.In the DDoS attack tool topic evolution study,it is found that most topics are not easily transferable among each other due to the restrictions on the use of DDoS attack tools,and the degree of confusion varies among a few topics.In the DDoS attack prediction study,it was found that the prediction accuracy was higher when fusing the two compared to predictions based on topics only or sentiment only.This thesis enriches and extends the research on DDoS attack tools and hacking forums.At the same time,this thesis proposes a new idea of fusing textual themes and sentiment for prediction,enriching the research on DDoS attack prediction based on hacker forums as a data source,and providing theoretical guidance and practical guidance for proactive response to DDoS attacks in advance.