# Machine Learning Based Intrusion Detection Systems for IoT Applications

**Authors:** Abhishek Verma, Virender Ranga  
**arXiv:** [2302.12452](https://arxiv.org/abs/2302.12452)  
**Date:** 2023-02-24  
**Summary page:** https://www.summarizepaper.com/en/paper/2302.12452v1/

## Key points

- The Internet of Things (IoT) is a popular research area with various applications
- IoT is exposed to cyber threats such as Denial of Service (DoS) attacks
- Machine learning classification algorithms are investigated for securing IoT against DoS attacks
- Comprehensive study on classifiers for anomaly-based intrusion detection systems (IDSs)
- Performance assessment using prominent metrics and validation methods
- Benchmarking classifiers using popular datasets CIDDS-001, UNSW-NB15, and NSL-KDD
- Statistical tests like Friedman and Nemenyi tests used to analyze differences among classifiers
- Evaluation of response time of classifiers on IoT-specific hardware using Raspberry Pi
- Methodology discussed for selecting the best classifier based on application requirements
- Motivation for developing IDSs using ensemble learning in IoT security research
- Suggestions for statistically assessing classifier performance
- Paper organized into sections providing an overview, discussing classification algorithms, explaining experimental design, presenting classifier performance and statistical test results, and concluding the paper

## Summary

The Internet of Things (IoT) has become a popular research area with various applications. However, it also exposes IoT to cyber threats such as Denial of Service (DoS) attacks. To address this issue, the authors investigate the use of machine learning classification algorithms for securing IoT against DoS attacks. They conduct a comprehensive study on classifiers that can advance the development of anomaly-based intrusion detection systems (IDSs). The performance of these classifiers is assessed using prominent metrics and validation methods. Popular datasets CIDDS-001, UNSW-NB15, and NSL-KDD are used for benchmarking the classifiers. Statistical tests like Friedman and Nemenyi tests are employed to analyze significant differences among classifiers. Additionally, the response time of classifiers on IoT-specific hardware is evaluated using Raspberry Pi. The authors also discuss a methodology for selecting the best classifier based on application requirements. This study aims to motivate IoT security researchers to develop IDSs using ensemble learning and suggest appropriate methods for statistically assessing classifier performance. In terms of organization, Section 2 provides an overview of recent works in defending IoT against DoS attacks. Section 3 discusses classification algorithms, including single classifiers and ensembles. The experimental design is explained in Section 4 while Section 5 discusses the performance of the classifiers and presents statistical test results. Finally, Section 6 concludes the paper. In summary, this paper contributes to the field by exploring machine learning techniques for securing IoT against DoS attacks and providing insights into selecting suitable classifiers for different application requirements.