Research On Social Spammer Detection Strategy Based On Semi-supervised Broad Learning

In recent years,Mobile Social Networks(MSN)have become an integral part of people's daily lives.Mobile social networks include a large number of social members who forward messages cooperatively.However,spammers post links to viruses and advertisements,or follow a large number of users,which produces many misleading messages in mobile social networks and seriously threatens the security of social users.In this dissertation,an adaptive social spammer detection(ASSD)model is proposed.In the process of using machine learning methods to train malicious user detection models,the cost of collecting a large amount of labeled social data with artificial labels or restrictive rules is high.A semi-supervised broad learning system with high detection accuracy is proposed for spammer detection.ASSD is trained with a small number of labeled patterns and a large number of unlabeled patterns.Compared with the traditional supervised learning method,it combines the similarity attributes between social users to build a high-precision malicious user detection model.Moreover,the time and energy required to label the identity of social members are reduced by applying ASSD.At the same time,the time complexity and computational complexity of the algorithm in the prediction process are low,and it can be deployed directly to the mobile terminal.Because social spammers frequently change their behavior to deceive the spammer detection model,an incremental learning method is designed to update the spammer detection model adaptively,without retraining.ASSD is compared with other supervised and semi-supervised machine learning methods using social data sets.Experimental results show that ASSD can build a high-accuracy model with a small number of labeled samples.Furthermore,it maintains effective performance without retraining to adapt to newly generated social data.Experiments on the Twitter dataset show that ASSD reaches a high accuracy with a small number of labeled patterns.Meanwhile,the classification adaptively updates to maintain the efficiency over time,rather than being cheated by spammers.