MALICIOUS CONTENT DETECTION IN SOCIAL NETWORKS USING HYBRID MACHINE LEARNING MODEL

Abstract

Social Networking platforms like Facebook, Twitter, Reddit, Weibo, Instagram and many more are the most popular and very easy to use medium for social connectivity, staying up to date with current events and news, relaxing in spare time, sharing opinions on many occurring events etc., Usage of these platforms have tremendously increased year over year say 9 to 12%. As of 2021 half of the percentage of people are using social media out of entire population [1]. With this much usage, if the entire information available in the Network is real and informative then it is really appreciated, but there is clear evidence that there are high chances of dissemination of malicious information for variety of reasons which creates negative impact on the society. So, detecting this type of content in the Social Network is very important research area. From past many years researchers have come up with different ideas to identify this type of information with Data Mining, Machine Learning, Deep Learning techniques. In this paper we propose a hybrid approach HCSTCM (Hybrid Cluster derived Sentiment based Topic Classification Model) to identify malicious content in the Social Network by deriving clusters, sentiments and topic information of the documents, later on using these features for supervised learning. Main aim of the paper is to identify the most dependent features which effect the malicious content without redundancy and improve the classification accuracy. The proposed method is validated with three social platform data.