Classification of Adversarial Attacks Using Ensemble Clustering Approach

As more business transactions and information services have been implemented via communication networks, both personal organization assets encounter a higher risk of attacks. To safeguard these, perimeter defence like NIDS (network-based intrusion detection system) can be effective for known intrusions. There has great deal attention within the joint community security data science to improve machine-learning based such that it becomes accurate adversarial attacks, where obfuscation techniques are applied disguise patterns intrusive traffics. The current research focuses on non-payload connections at TCP (transmission control protocol) stack level is applicable different network applications. In contrary wrapper method introduced with benchmark dataset, three new filter models proposed transform feature space without knowledge class labels. These ECT (ensemble clustering transformation) techniques, i.e., ECT-Subspace, ECT-Noise ECT-Combined, developed using concept ensemble generation strategies, random subspace, noise injection their combinations. Based empirical study published dataset four classification algorithms, usually outperform original other alternatives found in literature. This similarly summarized from first experiment basic legitimate direct second recognizing obfuscated addition, analysis algorithmic parameters, size noise, provided as guideline practical use.