TY - GEN
T1 - Recursive Feature Selection and Intrusion Classification in NSL-KDD Dataset Using Multiple Machine Learning Methods
AU - Mohanty, Subrat
AU - Agarwal, Mayank
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
PY - 2024/1/1
Y1 - 2024/1/1
N2 - IDS are critical components of modern computer networks, designed to detect and alert administrators of malicious activity. In order to detect network irregularities and keep data secure, it is critical to build an effective IDS that prevents unauthorized access to network resources. In this study, several machine learning classifiers were used to detect attacks in the NSL-KDD dataset. These classifiers included SVM, Naive Bayes, Random Forest, Decision Tree, and XGBoost. We have chosen 13 feature subsets using the recursive feature selection technique from the NSL-KDD dataset and used them to assess the model’s performance. Because the dimension of the data influences how well this IDS performs, the data was pre-processed, and superfluous attributes were deleted. The experimental results demonstrate that for all attack classes utilizing distinctive feature subsets, the accuracy of Decision Tree (DT), Nave Bayes (NB), Random Forest (RF), Linear Regression, XGBoost, AdaBoost, and Support Vector Machine (SVM) was over 95%. Overall, the performance of XGBoost in conjunction with recursive feature selection was the best.
AB - IDS are critical components of modern computer networks, designed to detect and alert administrators of malicious activity. In order to detect network irregularities and keep data secure, it is critical to build an effective IDS that prevents unauthorized access to network resources. In this study, several machine learning classifiers were used to detect attacks in the NSL-KDD dataset. These classifiers included SVM, Naive Bayes, Random Forest, Decision Tree, and XGBoost. We have chosen 13 feature subsets using the recursive feature selection technique from the NSL-KDD dataset and used them to assess the model’s performance. Because the dimension of the data influences how well this IDS performs, the data was pre-processed, and superfluous attributes were deleted. The experimental results demonstrate that for all attack classes utilizing distinctive feature subsets, the accuracy of Decision Tree (DT), Nave Bayes (NB), Random Forest (RF), Linear Regression, XGBoost, AdaBoost, and Support Vector Machine (SVM) was over 95%. Overall, the performance of XGBoost in conjunction with recursive feature selection was the best.
KW - IDS
KW - Machine Learning
KW - NSL-KDD dataset
KW - Recursive feature selection
UR - http://www.scopus.com/inward/record.url?scp=85190368109&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-56998-2_1
DO - 10.1007/978-3-031-56998-2_1
M3 - Conference contribution
AN - SCOPUS:85190368109
SN - 9783031569975
T3 - Communications in Computer and Information Science
SP - 3
EP - 14
BT - Computing, Communication and Learning - 2nd International Conference, CoCoLe 2023, Proceedings
A2 - Panda, Sanjaya Kumar
A2 - Rout, Rashmi Ranjan
A2 - Bisi, Manjubala
A2 - Sadam, Ravi Chandra
A2 - Li, Kuan-Ching
A2 - Piuri, Vincenzo
PB - Springer Science and Business Media Deutschland GmbH
T2 - 2nd International Conference on Computing, Communication, and Learning, CoCoLe 2023
Y2 - 29 August 2023 through 31 August 2023
ER -