Sistem Klasifikasi Trafik Jaringan menggunakan Algoritma XGBoost untuk mendeteksi SSH-Bruteforce pada Jaringan Komputer

Artefaktianto Ajiromadhon, Ir. Budi Bayu Murti, S.T., M.T.

2026 | Tugas Akhir | D4 TEKNOLOGI JARINGAN

Meningkatnya penggunaan jaringan komputer menyebabkan perlunya analisis trafik untuk membedakan aktivitas normal dan SSH-Bruteforce. Penelitian ini merancang dan mengimplementasikan sistem klasifikasi trafik jaringan berbasis Machine Learning menggunakan XGBoost untuk membedakan trafik Benign dan SSH-Bruteforce berdasarkan karakteristik network flow. Data diperoleh melalui capture trafik menggunakan CICFlowMeter, kemudian diproses melalui preprocessing dan feature engineering dengan 27 fitur. Dataset dibagi menjadi 70?ta latih dan 30?ta uji, sedangkan data latih diseimbangkan menggunakan SMOTE sehingga menghasilkan 402 data latih. Sistem diimplementasikan pada Ubuntu Server dengan CICFlowMeter, SQLite, dan web dashboard berbasis Flask untuk menyajikan hasil klasifikasi mendekati kondisi real-time. Evaluasi terhadap 150 data uji menghasilkan accuracy 99,33%, precision, recall, dan F1-score sebesar 0,99, serta ROC-AUC sebesar 0,9984. Pengujian menggunakan Hydra dan Ncrack menghasilkan 572 flow SSH, dengan 569 flow atau 99,48?rhasil diklasifikasikan secara langsung oleh model dan weighted average confidence sebesar 98,98%. Hasil tersebut menunjukkan bahwa XGBoost mampu mengklasifikasikan trafik Benign dan SSH-Bruteforce dengan tingkat kesalahan yang rendah serta dapat diintegrasikan dalam sistem klasifikasi trafik jaringan.

The increasing use of computer networks necessitates traffic analysis to distinguish normal activities from SSH-Bruteforce attacks. This study designs and implements a network traffic classification system based on Machine Learning using XGBoost to distinguish between Benign and SSH-Bruteforce traffic based on network flow characteristics. The data were obtained through traffic capture using CICFlowMeter and subsequently processed through preprocessing and feature engineering involving 27 features. The dataset was divided into 70% training data and 30% testing data, while the training data were balanced using SMOTE, resulting in 402 training data samples. The system was implemented on an Ubuntu Server using CICFlowMeter, SQLite, and a Flask-based web dashboard to present classification results in near-real-time. Evaluation using 150 test data samples achieved an accuracy of 99.33%, with precision, recall, and F1-score of 0.99, and a ROC-AUC of 0.9984. Testing using Hydra and Ncrack generated 572 SSH flows, of which 569 flows, or 99.48%, were successfully classified directly by the model, with a weighted average confidence of 98.98%. These results demonstrate that XGBoost is capable of classifying Benign and SSH-Bruteforce traffic with a low error rate and can be integrated into a network traffic classification system.

Kata Kunci : XGBoost, SSH Bruteforce, Machine Learning, CICFlowMeter, SMOTE, Keamanan Jaringan