# Building an Efficient Intrusion Detection System Based on Feature Selection and Ensemble Classifier

**Authors:** Yuyang Zhou, Guang Cheng, Shanqing Jiang, Mian Dai  
**arXiv:** [1904.01352](https://arxiv.org/abs/1904.01352)  
**Date:** 2019-04-02  
**Summary page:** https://www.summarizepaper.com/en/paper/1904.01352v4/

## Key points

- The paper proposes a new intrusion detection framework to address challenges faced by existing algorithms
- Three main issues highlighted: presence of redundant and irrelevant data, limited performance of individual classifiers, lack of adaptability to novel attacks
- Two-step approach: dimensionality reduction using CFS-BA algorithm, ensemble approach combining C4.5, RF, and Forest PA algorithms
- Voting technique used to combine probability distributions for attack recognition
- Framework evaluated using NSL-KDD, AWID, and CIC-IDS2017 datasets
- Experimental results show that the proposed method outperforms other approaches in various metrics
- Comprehensive solution incorporating feature selection and ensemble learning techniques

## Summary

The paper titled "Building an Efficient Intrusion Detection System Based on Feature Selection and Ensemble Classifier" proposes a new intrusion detection framework to address the challenges faced by existing intrusion detection algorithms. The authors highlight three main issues: the presence of redundant and irrelevant data in high-dimensional datasets, the limited performance of individual classifiers for different types of attacks, and the lack of adaptability of models to novel attacks. To overcome these challenges, the authors propose a two-step approach. In the first step, they introduce a heuristic algorithm called CFS-BA for dimensionality reduction. This algorithm selects an optimal subset of features based on their correlation, effectively reducing interference from redundant and irrelevant data. In the second step, the authors propose an ensemble approach that combines three different algorithms: C4.5, Random Forest (RF), and Forest by Penalizing Attributes (Forest PA). By leveraging the strengths of each algorithm, this ensemble approach improves the overall performance of attack detection. Furthermore, a voting technique is used to combine the probability distributions of the base learners for attack recognition which further enhances accuracy and reliability of intrusion detection. The proposed framework is evaluated using three datasets: NSL-KDD, AWID, and CIC-IDS2017. The experimental results demonstrate that the CFS-BA-Ensemble method outperforms other related approaches in terms of various metrics. Overall, this paper presents a comprehensive solution to enhance efficacy of intrusion detection systems by incorporating feature selection and ensemble learning techniques which shows promising results in improving performance and adaptability to novel attacks.