Today, IDS are unable to handle new and complex attacks effectively and are characterised by high false positives. To overcome these difficulties, there is a Multi-Layered Intrusion Detection System (ML-IDS) that includes a proposal to improve the detection efficiency by using the ensemble approach. Therefore our approach in this work proposes the use of layers of detection where each layer is optimised for a certain type of threat and the use of ensemble learning techniques like bagging and boosting in combining multiple machine learning algorithms for accurate predictions. This multilevel structure makes the system significantly more effective at identifying any type of intrusion, be it a known signature-based or a new anomaly. To measure the effectiveness of our proposed system we benchmark against existing datasets and single-layer IDS frameworks. They also show that our method increases the detection accuracy, decreases the number of false alarms, and generalises well to new types of attacks.