TTFSE-Two-Tier Feature Selection and ExtractionMachine Learning Model for Effective Network Attack Detection

Abstract

The Network Infringement Apperception System is vital tool to act against Infringements on computer networks and also protect the network from attacks by detecting and activating Rollback Mechanism to go back to safe state. This paper proposes a Two-Tier Feature Selection and Extraction Machine learning model, based on SelectKBest and Extra Tree Classifier for selecting, extracting and classifying the attack/normal instances in a network. This model encompasses two stages: The paramount tier is responsible for extracting top 40 features across 44 features in order to eliminate the features that have a less impact on detection of network infringement and the extracted features are used as input to succeeding prediction stage, here only 17 features which have high sway on detection of attack. This system uses Label Encoder to change categorical values of the dataset. By measuring its efficiency, several experiments are performed on a public dataset particularly on UNSW_NB15 dataset. The results shows TTFSE ML model has high performance, reduces the training time and is efficient for UNSW-NB15 Dataset.