Risk Assessment in Online Social Networks Through Client Activity Analysis using Machine Learning

Online Social Networks (OSN?s) are used to create a public or private pro?le that help in sharing information with other users and communicating with each other. These days OSN?s are increasingly susceptible to privacy assaults that affect both users and their connections. In such cases risk assessment rating is given for each account. The idea is that malicious user behavior is different from normal behavior, which should be considered risky. As such, a particular basic behavioral pattern can't be described that suits all behaviors of OSN users. However, we expect related people to continue to obey common standards with related behavioral patterns. This research study introduces a twophase risk assessment technique, leveraging machine learning to identify atypical user behaviors that deviate from established norms. The initial phase groups similar users called as Group Identification Phase, and the second phase develops behavior models for these groups (Risk Assessment Phase). This study is carried out based on two features (i. e., Group Identifiers (GI) and Behavioral Features (BF)). The goal of this two phase risk assessment technique is to find most likely group for a given data set. The study applies KNearest Neighbors KNN for classification, categorizing users based on behavioral traits. Three groups were created and were named as Randomized Features (RF), Smart Risky User (SMR), and More Smart Risky Users (MSRU). The effectiveness of our risk assessment approach is measured based on three output metrics i. e., F - measure, detection rate and false alarm rate. This approach demonstrated notable improvements in risk assessment, with higher accuracy in detection rates and reduced false alarms.