Improving the Feature Set in IoT Intrusion Detection Problem Based on FP-Growth Algorithm

This paper proposes and develops a new method to improve the feature set in IoT intrusion detection problem based on FP-Growth algorithm. Extracting a good feature set makes an important contribution to improving the accuracy of the IoT intrusion detection problem. Current studies on the detection of new IoT attacks focus on extracting independent features without considering the relationship between specific features and groups of features. Therefore, in the paper, we analyze and evaluate the original feature set to find the relationship between feature groups and add new features generated from relationships with a certain threshold to the original feature set to create the improved feature set. The detection of relationships reaching threshold is based on combination law mining technique according to FP-Growth algorithm. The improved feature set was tested with detection method based on convolution neural network with accuracy up to 98.2% and 1.45% improvement compared to the original feature set.