Malicious Activities Prediction Over Online Social Networking Using Ensemble Model

With the vast advancements in Information Technology, the emergence of Online Social Networking (OSN) has also hit its peak and captured the attention of the young generation people. The clone intends to replicate the users and inject massive malicious activities that pose a crucial security threat to the original user. However, the attackers also target this height of OSN utilization, explicitly creating the clones of the user's account. Various clone detection mechanisms are designed based on social-network activities. For instance, monitoring the occurrence of clone edges is done to restrict the generation of clone activities. However, this assumption is unsuitable for a real-time environment and works optimally during the simulation process. This research concentrates on modeling and efficient clone prediction and avoidance methods to help the social network activists and the victims enhance the clone prediction accuracy. This model does not rely on assumptions. Here, an ensemble Adaptive Random Subspace is used for classifying the clone victims with k-Nearest Neighbour (k-NN) as a base classifier. The weighted clone nodes are analysed using the weighted graph theory concept based on the classified results. When the weighted node's threshold value is higher, the trust establishment is terminated, and the clones are ranked and sorted in the higher place for termination. Thus, the victims are alert to the clone propagation over the online social networking end, and the validation is done using the MATLAB 2020a simulation environment. The model shows a better trade-off than existing approaches like Random Forest (RF), Naive Bayes (NB), and the standard graph model. Various performance metrics like True Positive Rate (TPR), False Alarm Rate (FAR), Recall, Precision, F-measure, and ROC and run time analysis are evaluated to show the significance of the model.