A Machine Learning Approach for Identifying Fake Accounts on Instagram

Abstract

Instagram is an increasingly popular social media platform in the digital social media ecosystem; however, the issue of fake and robotic accounts gaining momentum is dangerous and accordingly presents misinformation, scams, and identity abuse as significant threats. This study introduces an Instagram Fake Profile Detection system built on machine learning to detect suspicious users accounts with the help of behavioral, numerical, and profile-based features. The given system derives significant predictors, such as the ratio of followers to followed, the articles of the biography, the number of media, the structuring of the usernames, the presence of the profile picture, and privacy settings, and learns the trends typical of fake profiles. Several machine learning models are trained and compared, such as the Random Forest, SVM, ANN, GRU, LSTM, and Hybrid Deep Learning architecture, to increase the prediction accuracy. It adopts a Flask-based web interface enabling real-time classification and visualization of the model outputs of any instagram user input. The findings demonstrate that the reliability of fake account detection is enhanced due to the integration of traditional ML and deep learning networks. In general, this work would be a valuable addition to the body of research because it presents a functional, automatized, and scalable method of enhancing security and trust within the user space of Instagram.