Analysis of Clickstream Data to Detect Dangerous Chatbots in Networks

Abstract

With the great extent in the volume, velocity, and range of consumer statistics (e.g., user- generated data) in online social networks, there have been tried to graph new methods of amassing and examining such large data. For example, social bots have been used to function as automatic analytical offerings and furnish customers with elevated fantastic of service. However, malicious social bots have also been used to disseminate false facts (e.g., pretend news), and this can end result in real-world consequences. Therefore, detecting and doing away with malicious social bots in online social networks is crucial. The most current detection strategies of malicious social bots analyze the quantitative facets of their behavior. These facets are without problems imitated through social bots; thereby ensuing in low accuracy of the analysis. A novel approach of detecting malicious social bots, along with each aspect’s decision primarily based on the transition chance of clickstream sequences and semi-supervised clustering, is introduced in this paper. This approach now not solely analyzes transition likelihood of consumer conduct clickstreams however additionally considers the time function of behavior. Findings from our experiments on actual on line social community systems reveal that the detection accuracy for special kinds of malicious social bots with the aid of the detection technique of malicious social bots based totally on transition likelihood of person conduct clickstreams will increase by means of an common of 12.8%, in contrast to the detection approach based totally on quantitative evaluation of consumer behavior.