Social Media Fake Profile Classification: A New Machine Learning Approach

Abstract

Social media fake profile serves various illegal social activities. Therefore, detection and prevention of these profiles are essential. The current approaches based on machine learning (ML) are just considering social media user profile attributes by providing a strict classification. This paper provides a view to utilize a scoring-based fake profile classification technique to monitor the activities of user by using profile attributes and published content. This paper first includes a review to know the dataset to be used and the technique to obtain data from social media platform. Then based on social media user’s profile attribute-based ML model has been introduced to classify the fake and legitimate profiles. To train and validate the model, we have used five machine learning algorithms, namely artificial neural network (ANN), support vector machine (SVM), C4.5 decision tree, Bayes classifier, and k-nearest neighbor (k-NN). Here we have found ANN and SVM which is accurate classification technique as compared to others for this task. Finally, by updating the backpropagation neural network and a scoring method for profile a fake profile classification approach has been developed. The developed model is utilizing the content published by users and the basic profile information of public domain. The experiments have been carried out based on real twits and available profile attribute dataset in GitHub. The results are also compared with SVM and ANN algorithms. Based on the precision, recall, and F-score, the proposed technique outperforms as compared to other two implemented models and has been achieved up to 0.94 f-score.