Diagram Kontrol Multivariat Robust T2 Hotelling Berbasis Minimum Regularized Covariance Determinant Dengan Bootstrap Control Limit Untuk Deteksi Intrusi

Prasetya, Ichwanul Kahfi (2024) Diagram Kontrol Multivariat Robust T2 Hotelling Berbasis Minimum Regularized Covariance Determinant Dengan Bootstrap Control Limit Untuk Deteksi Intrusi. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003221026-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (2MB) | Request a copy

Abstract

Diagram kontrol T2 Hotelling merupakan salah satu diagram kontrol multivariat yang biasa digunakan untuk memonitor pergeseran proses mean. Diagram kontrol ini juga dapat diterapkan untuk mendeteksi outlier pada Intrusion Detection System (IDS). Deteksi intrusi umumnya dilakukan dengan mencocokkan pola network traffic normal dan mencari pola yang aneh sebagai anomali atau outlier. Karena banyaknya outlier, dataset IDS biasanya memiliki unknown distribution. Sayangnya, kelemahan diagram kontrol T2 Hotelling konvensional adalah diagram yang dibangun berdasarkan asumsi distribusi normal multivariat dan tidak sensitif untuk mendeteksi keberadaan mutiple outliers karena efek masking dan swamping, sehingga diagram tersebut perlu dikembangkan dan ditingkatkan. Estimator robust dapat digunakan untuk meningkatkan kinerja diagram kontrol dalam mendeteksi outlier. Salah satu penduga kuat terbaru adalah Minimum Regularized Covariance Determinant (MRCD) yang dapat digunakan untuk mendeteksi outlier terutama pada kumpulan data berdimensi tinggi. Untuk mengatasi masalah unknown distribution, pendekatan non-parametrik seperti metode bootstrap resampling dapat diterapkan untuk mengestimasi batas kontrol. Penelitian ini berfokus pada pengembangan diagram T2 Hotelling berbasis MRCD dengan Bootstrap Control Limit. Metode ini akan diterapkan pada studi simulasi dan diimplementasikan pada sistem deteksi intrusi dengan menggunakan dataset UNSW-NB15. Berdasarkan studi simulasi, T2 Hotelling berbasis MRCD memiliki performa yang lebih baik dalam mendeteksi outlier dibandingkan T2 Hotelling konvensional dan T2 Hotelling berbasis Fast-MCD. Pada penerapan data, T2 Hotelling berbasis MRCD juga memiliki performa yang lebih baik dalam mendeteksi intrusi meskipun memerlukan waktu eksekusi yang relatif lama.
=============================================================================================================================
Hotelling’s T2 control chart is one of the multivariate control charts is commonly used to monitor the mean shift process. This control chart can also be adopted to detecting outliers in Intrusion Detection System (IDS). Intrusion detection is generally done by matching normal network traffic patterns and looking for abnormal patterns as anomalies or outliers. Due to multiple outliers, the IDS datasets commonly have unknown distribution. Unfortunately, the drawback of conventional Hotelling’s T2 chart constructed under normal multivariate distribution assumption and it is not sensitive to detect the presence of multiple outliers due to masking and swamping effect, so the chart needs to be developed and enhanced. Robust estimator can be utilized to increase control chart’s performance to detecting outliers. One of the recent robust estimators is Minimum Regularized Covariance Determinant (MRCD) that can be used to detect outliers especially in high-dimensional data sets. To solve unknown distribution problem, non-parametric approach such as bootstrap resampling methods can be applied to estimate the approximate control limit. This research focused on developing Robust MRCD based Hotelling’s T2 chart with Bootstrap Control Limit. This method will be applied on simulation study and implemented in IDS by using UNSW_NB15 dataset. Based on simulation studies, MRCD based Hotelling’s T2 has better performance in detecting outliers compared to conventional Hotelling’s T2 and Fast-MCD based Hotelling’s T2. In data application, MRCD based Hotelling’s T2 also has better performance in detecting intrusions even though it requires a relatively long execution time.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Bootstrap, Deteksi Intrusi, Diagram Kontrol Multivariat, MRCD, T2 Hotelling, Intrusion Detection, Multivariate Control Chart, Hotelling’s T2
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD9980.5 Service industries--Quality control. Q Science
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Ichwanul Kahfi Prasetya
Date Deposited:	18 Feb 2024 13:39
Last Modified:	18 Feb 2024 13:39
URI:	http://repository.its.ac.id/id/eprint/107347

Actions (login required)

View Item