Research And Implementation Of A Spammer Detection System In Social Networks Based On User Profile

With the popularity of smart phones and the development of the mobile internet,social network plays more and more important role in people's daily life.Users browse information and share life on social networks.While enjoying the convenience brought by social networks,people are also affected by spam posted by malicious users.Hence,many scholars have conducted research on the detection of malicious users in social network.Based on the user's contents,behavior,social relationships and other characteristics,the existing researches mainly use Machine Learning,Deep Learning,Graph Theory to classify malicious users.However,user-generated contents are difficult to obtain because of user privacy protection.What's more,the features based contents are sparse and noisy.The features based on behavior are so easily to be evaded by malicious users,so it's impossible to detect advanced malicious users.It is difficult to compute the features based on social relationships,and the features based on social relationships always have high dimensionsTo address the above issues,this paper proposes and implements a malicious user detection model based on user profile.The main work and innovations of the paper are as follows:(1)A malicious user detection model in social networks based on user profile is proposed.Firstly,the model constructs a user profile labelling system,which contains user demographic profiling,user behavior profiling and user social profiling.Secondly,the model uses PBWL(PrefixSpan Based on Weight and Location)to mine user operating habits in user behavior profiling.Furthermore,the model uses the methods of statistical analysis to compute users' online periods,activity,average number of operations,most commonly used operations.As for the social profiling,the model uses WPRI(PageRank Combined with Interaction Rate)to compute social influence of users.Finally,the user profile is used as characteristics to detect malicious users.We have conducted a lot of experiments on the dataset from tagg.com,experimental results show that our proposed model improves the Precision compared with the Multi-level Dependency Model(MDM),which proves the effectiveness of the proposed model.(2)A malicious user detection system based on user profile is implemented.Firstly,we analyze the requirements of the system from three aspects:business,functional,and non-functional requirements analysis.We then design the overall architecture of the system and finally develop the modules such as data processing module,user profile generation module,user profile display module,and malicious user detection module.The process of malicious user detection is more transparent to monitors through the implementation of system.The system also demonstrates the availability and the high efficiency of the proposed model.