AURUM Mühendislik Sistemleri ve Mimarlık Dergisi

Yazarlar: Oğuz ATA, Khalid KADHİM

Konular:-

Anahtar Kelimeler:Machine Learning,Artificial Neural Network,Intrusion Detection Systems

Özet: Abstract Recently, it has become important to use advanced intrusion detection techniques to protect networks from the developing network attacks, which are becoming more complex and difficult to detect. For this reason, machine learning techniques have been employed in the Intrusion Detection Systems (IDS), so that, more complex features can be detected in the characteristics of the packets incoming to the network. As these techniques require training data, many datasets are collected for this purpose. Some of these datasets have known issues that limit the ability to apply intrusion detection systems built, based on these datasets, in real-life applications. In this study, the existing intrusion datasets are illustrated alongside with the known issues of each dataset, as well as, the existing intrusion detection systems that employ machine learning techniques and use these datasets, are discussed. As machine learning techniques extract different knowledge from different datasets, and each technique has different approaches to extract that knowledge, the performance of each technique is different from one dataset to another. The results of the discussed studies show the great potential of using machine learning techniques to implement IDS, where the Artificial Neural Networks (ANN) have shown the highest average performance, among other machine learning techniques.

---