A Hybrid Classification Approach of Network Attacks using Supervised and Unsupervised Learning

The increasing scale and sophistication of network attacks have become a major concern for organizations around the world. As result, there is an demand effective accurate classification to enhance cyber security measures. Most existing schemes assume that available training data labeled; is, based on supervised learning. However, this not always case since real expected be unlabeled. In paper, issue tackled by proposing hybrid approach combines both unsupervised learning build predictive model classifying attacks. First, used label in dataset. Then, different machine algorithms are utilized classify with labels obtained from first step compare results ground truth labels. Moreover, unbalanced dataset addressed using over-sampling under-sampling techniques. Several experiments been conducted, NSL-KDD dataset, evaluate efficiency proposed demonstrate accuracy our comparable methods all labeled.