EVALUATION OF CYBER SECURITY DATA SET CHARACTERISTICS FOR THEIR APPLICABILITY TO MACHINE LEARNING ALGORITHM DETECTING CYBER SECURITY ANOMALIES

Abstract

Artificial Intelligence algorithms play an important role in network security and attack detection, and in some cases can provide better results than classical detection tools such as Snort or Suricata. In this sense, the aim of this study is to evaluate the features of various mature machine learning algorithms frequently used in IDS scenarios. To do this, first consider splitting the data security network configuration and separating its data into different groups. Using this classification, this task is to determine which neural network model (multilayer or recurrent), activation function, and learning algorithm produces more results based on the dataset output. Finally, the results are used to determine which data in the network security dataset are more relevant and representative for access detection, as well as the optimal machine learning algorithm configuration to reduce physical computation.