Host Intrusion Detection System Using Novel Predefined Signature Patterns by Comparing Random Forest over Decision Tree Algorithm

Abstract

The objective of the work is to identify anomaly intrusion detection in a user network environment using information accessing and retrieval from homogeneous network databases. Here machine learning algorithms namely Random Forest and Decision Tree are been used to categorize passive and active attack. To accomplish focused accuracy, sample size of n=10 in Random Forest and n=10 in Decision Tree was repeated for 20 intervals for well-organized and precise investigation on categorized images with G power in 80% and threshold 0.05%, CI 95% mean and standard deviation. The existing works proves the sequential implemented in focus to intrusion detection, while comparing Random Forest and Decision Tree has classified and predicted the values from the network intrusion to generate accuracy with Random Forest has higher accuracy (76.37%) compared to Decision Tree accuracy (71.57%) with a significance of P<0.001 (2-tailed). Prediction in identifying anomaly intrusion detection systems shows that Random Forest has higher accuracy over Decision Tree.