Hybrid Ensemble Learning Sistem Keamanan Jaringan Untuk Meningkatkan Performa Deteksi Anomali
Gespeichert in:
| Titel: | Hybrid Ensemble Learning Sistem Keamanan Jaringan Untuk Meningkatkan Performa Deteksi Anomali |
|---|---|
| Autoren: | Irawan, Rony Heri, Irawan, Rony Heri Irawan, Nico Adi Saputra, Umi Mahdiyah |
| Quelle: | Nusantara of Engineering (NOE); Vol. 8 No. 02 (2025): Volume 8 Nomor 2-2025; 361 – 369 Nusantara of Engineering (NOE); Vol 8 No 02 (2025): Volume 8 Nomor 2-2025; 361 – 369 |
| Verlagsinformationen: | Universitas Nusantara PGRI Kediri, 2025. |
| Publikationsjahr: | 2025 |
| Schlagwörter: | Deteksi Anomali, Hybrid Ensemble, Keamanan Siber, Serangan Siber, Network Intrusion Detection Sistem, Anomaly Detection, Cyber Attack, Cyber Security, Hybrid Ensemble, Network Intrusion Detection System |
| Beschreibung: | Serangan siber seperti zero-day attacks dan APT menjadi tantangan serius bagi sistem deteksi intrusi jaringan, terutama yang masih mengandalkan metode berbasis tanda tangan. Penelitian ini bertujuan merancang sistem deteksi anomali jaringan berbasis hybrid ensemble learning dengan menggabungkan algoritma Isolation Forest, K-Means, dan Random Forest menggunakan metode majority voting. Proses penelitian meliputi preprocessing data, pelatihan dan evaluasi model menggunakan dataset publik CSE-CIC-IDS2018. Evaluasi dilakukan dengan metrik akurasi, precision, recall, F1-score, dan AUC. Hasil menunjukkan bahwa pendekatan hybrid ini meningkatkan akurasi deteksi hingga 99,9% dan menurunkan false positive secara signifikan dibanding pendekatan tunggal. Sistem yang diusulkan terbukti lebih adaptif dan efisien dalam mengidentifikasi berbagai pola serangan siber, serta memberikan kontribusi terhadap pengembangan teknologi keamanan jaringan yang lebih andal. Cyber attacks such as zero-day attacks and APTs are serious challenges for network intrusion detection systems, especially those that still rely on signature-based methods. This study aims to design a hybrid ensemble learning-based network anomaly detection system by combining the Isolation Forest, K-Means, and Random Forest algorithms using the majority voting method. The research process includes data preprocessing, training, and model evaluation using the CSE-CIC-IDS2018 public dataset. Evaluation is carried out using accuracy, precision, recall, F1-score, and AUC metrics. The results show that this hybrid approach improves detection accuracy by up to 99.9% and significantly reduces false positives compared to a single approach. The proposed system is proven to be more adaptive and efficient in identifying various cyber attack patterns, as well as contributing to the development of more reliable network security technology. |
| Publikationsart: | Article |
| Dateibeschreibung: | application/pdf |
| Sprache: | English |
| ISSN: | 2355-6684 2776-6640 |
| DOI: | 10.29407/noe.v8i02.25617 |
| Zugangs-URL: | https://ojs.unpkediri.ac.id/index.php/noe/article/view/25617 |
| Rights: | CC BY SA |
| Dokumentencode: | edsair.9e0ed7a5532b..889a7626a891d23c3d17fb8c1f18cb20 |
| Datenbank: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | Serangan siber seperti zero-day attacks dan APT menjadi tantangan serius bagi sistem deteksi intrusi jaringan, terutama yang masih mengandalkan metode berbasis tanda tangan. Penelitian ini bertujuan merancang sistem deteksi anomali jaringan berbasis hybrid ensemble learning dengan menggabungkan algoritma Isolation Forest, K-Means, dan Random Forest menggunakan metode majority voting. Proses penelitian meliputi preprocessing data, pelatihan dan evaluasi model menggunakan dataset publik CSE-CIC-IDS2018. Evaluasi dilakukan dengan metrik akurasi, precision, recall, F1-score, dan AUC. Hasil menunjukkan bahwa pendekatan hybrid ini meningkatkan akurasi deteksi hingga 99,9% dan menurunkan false positive secara signifikan dibanding pendekatan tunggal. Sistem yang diusulkan terbukti lebih adaptif dan efisien dalam mengidentifikasi berbagai pola serangan siber, serta memberikan kontribusi terhadap pengembangan teknologi keamanan jaringan yang lebih andal.<br />Cyber attacks such as zero-day attacks and APTs are serious challenges for network intrusion detection systems, especially those that still rely on signature-based methods. This study aims to design a hybrid ensemble learning-based network anomaly detection system by combining the Isolation Forest, K-Means, and Random Forest algorithms using the majority voting method. The research process includes data preprocessing, training, and model evaluation using the CSE-CIC-IDS2018 public dataset. Evaluation is carried out using accuracy, precision, recall, F1-score, and AUC metrics. The results show that this hybrid approach improves detection accuracy by up to 99.9% and significantly reduces false positives compared to a single approach. The proposed system is proven to be more adaptive and efficient in identifying various cyber attack patterns, as well as contributing to the development of more reliable network security technology. |
|---|---|
| ISSN: | 23556684 27766640 |
| DOI: | 10.29407/noe.v8i02.25617 |