An Effective Approach to Classify Abnormal Network Traffic Activities using Wavelet Transform

Abstract

Understanding network activities has become the most significant task in network security due to the rapid growth of the Internet and mobile devices usages. To protect our computing infrastructures and personal data from network intruders or attacks, identifying abnormal activities is critical. Extracting features from network traffic data is considered as an essential task to be performed because it affects the overall performances to identify the activities accurately. Although researchers proposed several approaches, they mainly focused on identifying the best possible technique to detect abnormal network activities. Only a few studies considered utilizing feature extraction techniques. In this paper, we introduced a new approach, with which an integrative information feature set is determined to identify abnormal network activities using wavelet transformation. Instead of extracting features by attributes, the approach uses all attributes information to extract features and to design a reliable learning model to detect abnormal activities by reducing false positives. Two machine learning techniques, Logistic Regression (LR) and Naive Bayes, are utilized to show the effectiveness of the approach. A visualization method is also used to emphasize our approach. As a result, we found that our proposed approach produces a better performance result with less computational time in detecting abnormal network activities.