Remote Access Trojans (RATs) have gathered reasonable attention in the evolving realm of cybersecurity due to their stealthy characteristics and the capacity to cause significant privacy and security infringements. This research explores proactive security methods using machine learning against Android RAT attacks by investigating the network behavior based analysis method to build a reliable RAT detection system. The system can deep inspect network traffic and classify Android RAT traffic using the ensemble learning methods. Eight different types of RAT traffic data are included in the training dataset to train various machine learning models. Experiments in the research show that the ensemble learning models have high accuracy in discriminating the RAT traffic from benign traffic with a AUC score of 0.99. The study contributes novel data pre-processing technique, identification of key features for detecting RAT vulnerability, and an ensemble learning based approach for autonomous RAT detection.