Abstract
For the protection of digital infrastructures, strong and flexible security structures are necessary due to the fast rise in cyber threats, such as spyware, phishing emails, and distributed denial-of-service (DDoS) attacks. By combining the advantages of several techniques, ensemble machine learning (EML) has become a potent paradigm to improve cyber defense by increasing detection accuracy and resistance against changing attack vectors. In order to successfully detect and prevent network intrusions, this research investigates an ensemble strategy that makes use of K-Nearest Neighbors (KNN), Long Short-Term Memory (LSTM) networks, and Multi-Layer Perception (MLP) models. MLP offers nonlinear feature training for complicated threat landscapes, LSTM is excellent at identifying sequential relationships in network data, and KNN offers effective recognition of patterns for static attack signatures. By combining these models, temporal and geographical features are exploited, lowering false positives and improving prediction accuracy. Recent benchmark datasets, such as CIC-DDoS2019, are used to assess performance in a variety of attack scenarios, offering a thorough understanding of practical application. The suggested ensemble performs noticeably better than individual models in accuracy, precision, and recall, according to experimental data, making it a viable instrument for proactive cyber defense tactics. This study emphasizes how ensemble learning may improve cybersecurity and network resilience in a revolutionary way.
Keywords: Cybersecurity, Distributed Denial-Of-Service, Ensemble Machine Learning, K-Nearest Neighbors, Long Short-Term Memory Networks, Multi-Layer Perception.