Research On Fraud Detection Methods In Social Network Marketin

In recent years,online social networks(OSNs)have gained a lot of attention as a type of web service.OSNs such as Twitter and Weibo have attracted a large number of users,changing the way people live and entertain themselves.Compared to traditional media,such as one-way broadcasting,printing,and television,social network marketing is more diversified and also more trusted by users.The growing number of personal and business users in social media highlights the importance of social media in commercial applications,especially in the field of marketing.However,due to the openness of usergenerated networks,some users engage in marketing fraud attacks under the inducement of commercial interests.This paper uses time-series behavior modeling and multi-task text pre-training models as technical means to detect marketing fraud in social networks.The main research content and contributions of this paper include:1.With the development of social networks,word-of-mouth information can be used for marketing to social network users.However,it is highly susceptible to false information attacks,in which attackers post false information or comments to mislead users for commercial gain.Recently,some new fields have also been targeted by false review attacks due to the development of social networks,and these fields have small data sizes,making it difficult to train effective models.To address this problem,this paper proposes a novel false review detection model based on generative adversarial networks(GANs)and domain adaptive learning,which includes a text feature encoder,a domain classifier,and a false review detector.First,the feature extractor uses a text pre-training model to encode rich information in the review text.Secondly,the domain classifier is implemented by a neural network discriminator,which can extract common features.Thirdly,the false review detector is used to identify false reviews.Finally,the min-max game between the domain classifier and the false review detector realizes transfer learning and enhances the overall generalization ability of the model.Extensive experiments show that our model has high detection accuracy for false review attacks across domains in cold-start situations.2.User-Review Social Networks(URSNs)now become the targets of Sybil attacks,where fake reviews are posted by attackers to raise the reputation of listed services or products.Unlike previous fake accounts on Twitter or Weibo,Sybil attackers on URSNs are also genuine users in many cases,which presents a challenge for the existing fake review detection system.Hence,the main purpose of detecting Sybil attackers is to profile abnormal behaviors of users,and we propose a novel Sybil detection model named MUSH to extract long-term features of user behaviors on URSNs.First,aiming to measure the uncertainty of user behaviors,four entropy-based preference models are designed to quantify user preferences,including catering preferences,price preferences,word-of-mouth preferences,and rating preferences.Second,in order to extract temporal logic features,a novel multi-stimuli Hawkes process is introduced by combining external incentives and internal incentives,which can detect abnormal event sequences of posted reviews.This approach is quite different from previous solutions which mostly use direct/indirect graph models or user-related features for detection.Finally,by integrating entropy-based preferences with temporal logic features,a smart Sybil detection model is proposed based on a binary classification approach.Extensive experimental results indicate that MUSH can effectively detect Sybil attackers with a high detection accuracy.By comparing with other approaches,MUSH is quite suitable for adversarial network environments,which could provide better security services to social networks.3.With the expansion of social networks,hashtags are increasingly popular for search and recommendation as highlighted topics of text microblogs.However,since hashtags could be casually and unrestrictedly created by users,they are highly susceptible to hashtag hijack attacks,especially when associated with the new commercial promotions emerged on social networks.To address this issue,previous efforts either target only a few unique hijacks or extract features to match the content of the microblog to the hashtag.However,these approaches fail to distinguish between innocent hashtag usage by normal users and actual hashtag hijack by attackers.In this paper,we propose Sec TAG,a multi-task learning framework to fast detect the truly hostile attackers by revealing the behavioral intentions of users.Our Sec TAG consists of a three-subtask hijack detection model,a dynamic weighting scheme and a hijack risk assessment scheme.First,based on both the deep learning model of BERT and LSTM,we design the three-subtask hijack detection model,which identifies the microblog whose content mismatches with the hashtag,separates out the non-malicious use of hashtags and finally detects hashtag hijack attacks.Second,we develop a dynamic weighting scheme for optimizing multi-task loss function of our three-subtask detection model.Finally,we construct the hijack risk assessment scheme to compute the risk level of hashtags,which has been validated effectively on the real-world dataset.Extensive experiments justify that Sec TAG detects hashtag hijack attacks accurately.