A Membership Function Detection of Low for Intrusion and Anomaly Frequency Attacks

The ultimate objective of intrusion detection problem is to identify surprising intrusions that compromise networks. Determining intrusions through the application of classifiers or detection algorithm requires, finding similarity as one of the important operations. This paper brings to the discussion a membership function that can be used for the learning process to attain better accuracies for low-frequency attack classes in the given dataset. Two membership functions are proposed in this work for unsupervised learning. The first one is utilized for prior learning and the second one is utilized for post-learning. The learning process is an un-supervised technique that aims at dimensionality transformation.