TY - GEN
T1 - Boosting Accuracy in Intrusion Detection Systems
T2 - 2nd IEEE International Conference on Networks, Multimedia and Information Technology, NMITCON 2024
AU - Suriya Prakash, J.
AU - Guntupalli, Chandra Haasitha
AU - Narasani, Snehitha
AU - Srinidhi, N. N.
AU - Kiran, S.
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Accurate classification of diverse attack types remains a challenge in network intrusion detection. This study employs Linear Discriminant Analysis (LDA) as a preprocessing step to improve the classification accuracy of machine learning algorithms applied to network traffic data. Using the CICIDS2017 dataset, popular classifiers such as Decision Trees, Naive Bayes, SVM, Logistic Regression, K-nearest Neighbors, and Random Forest are evaluated. Experiments are conducted with varying test sizes (0.2, 0.5, 0.7) to assess the generalizability of results. The findings indicate that LDA significantly enhances classification accuracy, precision, and recall for numerous classifiers compared to scenarios without LDA. Notable improvements are observed across different algorithms and test size configurations. Additionally, scenarios where classifiers without LDA achieve comparable results are identified, providing insights for algorithm selection in practical applications. This research contributes to optimizing intrusion detection systems and cybersecurity applications by offering a comprehensive understanding of LDA's role in enhancing algorithmic performance on the CICIDS2017 dataset. https://github.com/haasi003/Quantitative-overview-ofimproving-cybersecurity-classification-algorithms-with-LDA
AB - Accurate classification of diverse attack types remains a challenge in network intrusion detection. This study employs Linear Discriminant Analysis (LDA) as a preprocessing step to improve the classification accuracy of machine learning algorithms applied to network traffic data. Using the CICIDS2017 dataset, popular classifiers such as Decision Trees, Naive Bayes, SVM, Logistic Regression, K-nearest Neighbors, and Random Forest are evaluated. Experiments are conducted with varying test sizes (0.2, 0.5, 0.7) to assess the generalizability of results. The findings indicate that LDA significantly enhances classification accuracy, precision, and recall for numerous classifiers compared to scenarios without LDA. Notable improvements are observed across different algorithms and test size configurations. Additionally, scenarios where classifiers without LDA achieve comparable results are identified, providing insights for algorithm selection in practical applications. This research contributes to optimizing intrusion detection systems and cybersecurity applications by offering a comprehensive understanding of LDA's role in enhancing algorithmic performance on the CICIDS2017 dataset. https://github.com/haasi003/Quantitative-overview-ofimproving-cybersecurity-classification-algorithms-with-LDA
UR - https://www.scopus.com/pages/publications/85207417424
UR - https://www.scopus.com/pages/publications/85207417424#tab=citedBy
U2 - 10.1109/NMITCON62075.2024.10698893
DO - 10.1109/NMITCON62075.2024.10698893
M3 - Conference contribution
AN - SCOPUS:85207417424
T3 - 2nd IEEE International Conference on Networks, Multimedia and Information Technology, NMITCON 2024
BT - 2nd IEEE International Conference on Networks, Multimedia and Information Technology, NMITCON 2024
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 9 August 2024 through 10 August 2024
ER -