KLASIFIKASI SERANGAN PADA DATA LOG JARINGAN DENGAN SELEKSI FITUR INFORMATION GAIN PADA ALGORITMA RANDOM FOREST DALAM INTRUSION DETECTION SYSTEM (IDS)

Wijaya, Edwinpras (2024) KLASIFIKASI SERANGAN PADA DATA LOG JARINGAN DENGAN SELEKSI FITUR INFORMATION GAIN PADA ALGORITMA RANDOM FOREST DALAM INTRUSION DETECTION SYSTEM (IDS). Other thesis, UPN Veteran Yogyakarta.

	Text ABSTRAK.pdf Download (173kB)
	Text COVER.pdf Download (168kB)
	Text DAFTAR ISI.pdf Download (112kB)
	Text DAFTAR PUSTAKA.pdf Download (189kB)
	Text PENGESAHAN PEMBIMBING.pdf Download (212kB)
	Text PENGESAHAN PENGUJI.pdf Download (241kB)
	Text SKRIPSI FULL_EDWINPRAS WIJAYA_123200074.pdf Restricted to Repository staff only Download (2MB)

Abstract

vi
ABSTRAK
Intrusion Detection System (IDS) merupakan alat penting dalam keamanan jaringan
yang berfungsi untuk memantau dan mendeteksi serangan. IDS menjadi krusial dalam
mengidentifikasi serangan yang dapat melewati langkah-langkah keamanan tradisional
seperti firewall dan antivirus. Saat ini, penerapan machine learning telah banyak dilakukan
untuk meningkatkan kinerja deteksi intrusi pada data log sistem jaringan. Namun, beberapa
metode masih menghadapi keterbatasan seperti tingginya angka positif palsu, rendahnya
akurasi klasifikasi, dan rendahnya angka positif sebenarnya. Selain itu, fitur-fitur asing serta
besarnya jumlah data yang perlu diperiksa oleh IDS, membuat proses analisis menjadi sulit
bahkan dengan bantuan komputer. Oleh karena itu, pada penelitian ini akan menerapkan
seleksi fitur Information Gain dengan 3 nilai threshold berbeda yang akan dievaluasi untuk
menemukan nilai threshold terbaik pada seleksi fitur untuk Random Forest pada case
Intrusion Detection System (IDS).
Metodologi dari penelitian ini akan melibatkan beberapa tahap. Pertama, dilakukan
pengumpulan data, penelitian ini menggunakan dataset NSL-KDD, dimana data berupa log
jaringan yang sudah di proses. Lalu akan dilakukan preprocessing berupa class label
mapping, label encoding, splitting data, information gain, dan standardization. Lalu
pelatihan dan evaluasi model akan dilakukan. Data yang sudah melalui preprocessing akan
digunakan untuk melatih model, dan hasilnya akan dievaluasi dengan metrik akurasi, presisi,
recall dan f1-score. Dalam proses ini, akan dilakukan pemodelan untuk model multiclass
dan model binary class, dimana masing – masing model akan dibangun dengan 3 threshold
yang berbeda pada seleksi fitur information gain-nya.
Hasil dari pengujian pada model yang dibangun menunjukan hasil positif. Dengan
menguji 3 threshold berbeda (standar deviasi, 0,05, dan 0,07), model multiclass dengan
threshold 0,05 mencapai akurasi tertinggi sebesar 99,95%, dengan precision, recall, dan F1-
score yang berkisar antara 99,70% hingga 99,99% di seluruh kelas. Pada model binary class,
penggunaan threshold standar deviasi menghasilkan akurasi sebesar 99,88%, dengan
precision sebesar 99,85%, recall 99,93%, dan F1-score 99,89% untuk kelas "Attacking".
Namun, masih terdapat misklasifikasi pada kelas Probe, yang disebabkan oleh kemiripan
data kelas Probe dengan kelas Normal. Hasil pengujian ini menunjukkan bahwa penerapan
seleksi fitur Information Gain dengan threshold yang sesuai dapat meningkatkan efektivitas
deteksi serangan pada sistem IDS dengan algoritma Random Forest.
Kata Kunci: Intrusion Detection System, Information Gain, Random Forest, seleksi fitur,
threshold.
vii
ABSTRACT
An Intrusion Detection System (IDS) is an important tool in network security that
functions to monitor and detect attacks. IDS can be crucial in identifying attacks that can
bypass traditional security measures such as firewalls or an antivirus software. Currently,
the application of machine learning has been implemented to further enhance the
performance of intrusion detection in network system log data. However, most of the
methods used still face limitations such as a high rate of false positives, low classification
accuracy, and a low rate of true positives. In addition, the foreign features and the large
amount of data that need to be examined by the IDS make the analysis process difficult, even
with the help of computers. Therefore, this study will apply Information Gain feature
selection with 3 different threshold values that will be evaluated to find the best threshold
value for feature selection in Random Forest for the Intrusion Detection System case. (IDS).
The methodology of this research will involve several stages. First, data collection was
carried out; this research uses the NSL-KDD dataset, which consists of processed network
log data. Then, the data will goes to the preprocessing process, which consist of class label
mapping, label encoding, data splitting, information gain, and standardization. After that,
model training and evaluation will be conducted. The data that has been preprocessed will
be used to train the model, and the results will be evaluated using accuracy, precision, recall,
and F1-score metrics. In this process, modeling will be conducted for both multiclass and
binary class models, where each model will be built with three different thresholds based on
information gain for feature selection.
The test results from the constructed models shows positive outcomes. By testing three
different thresholds (standard deviation, 0.05, and 0.07), the multiclass model with a
threshold of 0.05 achieved the highest accuracy of 99.95%, with precision, recall, and F1-
score ranging from 99.70% to 99.99% across all classes. In the binary class model, using
the standard deviation threshold resulted in an accuracy of 99.88%, with precision at
99.85%, recall at 99.93%, and an F1-score of 99.89% for the "Attacking" class. However,
there is still misclassification in the Probe class, which is caused by the similarity of the
Probe class data to the Normal class. The results of this test indicate that the application of
Information Gain feature selection with an appropriate threshold can enhance the
effectiveness of attack detection in IDS systems using the Random Forest algorithm.
Keywords: Intrusion Detection System, Information Gain, Random Forest, feature selection,
threshold.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Intrusion Detection System, Information Gain, Random Forest, feature selection, threshold.
Subjects:	Z Bibliography. Library Science. Information Resources > ZA Information resources
Divisions:	Faculty of Engineering, Science and Mathematics > School of Engineering Sciences
Depositing User:	Eko Yuli
Date Deposited:	24 Sep 2024 01:16
Last Modified:	24 Sep 2024 01:16
URI:	http://eprints.upnyk.ac.id/id/eprint/41261

Actions (login required)

View Item