Claim Missing Document
Check
Articles

Model Hybrid Random Forest dan Information Gain untuk meningkatkan Performa Algoritma Machine Learning pada Deteksi Malicious Software Fauzi Adi Rafrastara; Wildanil Ghozi; Ramadhan Rakhmat Sani; L. Budi Handoko
Jurnal Informatika dan Rekayasa Perangkat Lunak Vol. 6 No. 2 (2024): September
Publisher : Universitas Wahid Hasyim

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

The evolution of malware, or malicious software, has raised increasing concerns, targeting not only computers but also other devices like smartphones. Malware is no longer just monomorphic but has evolved into polymorphic, metamorphic, and oligomorphic forms. With this massive development, conventional antivirus software is becoming less effective at countering it. This is due to malware's ability to propagate itself using different fingerprint and behavioral patterns. Therefore, an intelligent machine learning-based antivirus is needed, capable of detecting malware based on behavior rather than fingerprints. This research focuses on the implementation of a machine learning model for malware detection using ensemble algorithms and feature selection to achieve optimal performance. The ensemble algorithm used is Random Forest, evaluated and compared with k-Nearest Neighbor and Decision Tree as state-of-the-art methods. To enhance classification performance in terms of processing speed, the feature selection method applied is Information Gain, with 22 features. The highest results were achieved using the Random Forest algorithm and Information Gain feature selection method, reaching a score of 99.0% for accuracy and F1-Score. By reducing the number of features, processing speed can be increased by almost fivefold.
Deteksi Serangan Denial of Service (DoS) dan Spoofing pada Internet of Vehicles menggunakan Algoritma K-Nearest Neighbor (KNN) Wildanil Ghozi; Fauzi Adi Rafrastara; Ramadhan Rakhmat Sani; Abdussalam Abdussalam
Jurnal Informatika dan Rekayasa Perangkat Lunak Vol. 6 No. 2 (2024): September
Publisher : Universitas Wahid Hasyim

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

The implementation of Internet of Things (IoT) technology in motor vehicles has been increasing over time and is known as the Internet of Vehicles (IoV). IoV is becoming more essential to society as it provides comfort, safety, and efficiency in driving. Unfortunately, the use of internet technology in IoV brings the potential for cyber-attacks, such as Denial of Service (DoS) and Spoofing. Intrusion Detection Systems in IoV have not yet fully matured, as this technology is still relatively new. Therefore, the potential threats and their significant impact make research on this topic urgently needed. This study aims to evaluate the performance of the k-Nearest Neighbor (kNN) classification algorithm in detecting cyber-attacks on IoV. The predicted classes in this study consist of six categories: Benign, DoS, Gas-Spoofing, Steering Wheel-Spoofing, Speed-Spoofing, and RPM-Spoofing. These two types of attacks on IoV (DoS and Spoofing) pose risks to the operational safety of vehicles, which can endanger drivers and other road users. The dataset used is a public dataset called CIC IoV2024. The performance of the kNN algorithm is also compared to three other state-of-the-art algorithms, including Naïve Bayes, Deep Neural Network, and Random Forest. The results show that k-Nearest Neighbor (kNN) achieved the best performance with a score of 98.7% for both accuracy and F1-Score metrics. kNN outperformed Naïve Bayes, which ranked second with a score of 98.1% accuracy and 98.0% F1-Score. Thus, the kNN algorithm can be recommended as a classifier in the development of an intrusion detection system for IoV
Edukasi Perlindungan Privasi dan Antisipasi Kejahatan Siber di Kalangan Pelajar pada SMA Mardisiswa Semarang Ramadhan Rakhmat Sani; Wildanil Ghozi; Fauzi Adi Rafrastara; Yani Parti Astuti; Solichul Huda; Eko Hari Rachmawanto; Agus Winarno
ABDIMASKU : JURNAL PENGABDIAN MASYARAKAT Vol 9, No 3 (2026): SEPTEMBER 2026
Publisher : LPPM UNIVERSITAS DIAN NUSWANTORO

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.62411/ja.v9i3.3408

Abstract

Pelajar sekolah menengah atas merupakan kelompok pengguna internet dengan intensitas tinggi, menjadikan mereka rentan terhadap ancaman siber jika tidak dibekali dengan literasi digital yang memadai. Kurangnya pemahaman mengenai nilai data pribadi menyebabkan maraknya kasus penyalahgunaan data dan penipuan daring di kalangan remaja. Kegiatan pengabdian kepada masyarakat ini bertujuan untuk memberikan edukasi mengenai perlindungan privasi dan langkah antisipasi kejahatan siber kepada siswa SMA Mardisiswa Semarang. Metode pelaksanaan terdiri dari tiga tahapan utama, yaitu pre-test, pemaparan materi interaktif, dan post-test. Evaluasi pemahaman kognitif siswa diukur menggunakan platform daring Wayground.com. Hasil kegiatan menunjukkan bahwa pada tahap pre-test, rata-rata persentase jawaban benar dari 26 peserta adalah 79% dari total 15 soal. Setelah edukasi diberikan, hasil post-test yang diikuti oleh 11 siswa karena kendala teknis daya baterai gawai dan waktu pelajaran yang telah usai menunjukkan peningkatan rata-rata jawaban benar menjadi 90%, dengan 3 siswa di antaranya berhasil menjawab benar seluruh pertanyaan. Peningkatan ini mengindikasikan bahwa intervensi edukasi yang diberikan terbukti efektif dalam meningkatkan kesadaran siswa untuk melindungi privasi dan mengantisipasi berbagai bentuk kejahatan siber di dunia digital.
Improving Intrusion Detection under Extreme Class Imbalance through Class Representation Restructuring and SMOTE Ghozi, Wildanil; Mohamed Yassin, Warusia; Rakhmat Sani, Ramadhan; Rafrastara, Fauzi Adi
ELKHA Vol. 18 No.2 October 2026
Publisher : Faculty of Engineering, Universitas Tanjungpura

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26418/elkha.v18i2.107677

Abstract

Class imbalance is a major challenge in multi-class network intrusion detection, particularly when extremely underrepresented attack categories provide insufficient samples for effective model learning. Such imbalance can cause machine learning models to favor majority classes, resulting in high overall accuracy but poor detection of minority attacks. This study proposes a data-level imbalance handling strategy that restructures the representation of extremely sparse attack classes before applying the Synthetic Minority Over-sampling Technique (SMOTE). Four experimental conditions were evaluated on the UNSW-NB15 dataset using Decision Tree, Random Forest, and XGBoost classifiers to examine the effects of class representation restructuring and SMOTE separately. The evaluation employed accuracy, macro-precision, macro-recall, and macro-F1-score, complemented by class-wise analysis. The results show that, for XGBoost, class representation restructuring without SMOTE increases macro-recall from 54.44% to 68.56% and macro-F1-score from 57.44% to 68.24%, while accuracy remains nearly unchanged. When SMOTE is subsequently applied to the restructured representation, macro-recall further increases to 78.60% and macro-F1-score to 69.05%. Compared with the original non-SMOTE condition, the combined strategy improves macro-recall by 24.16% and macro-F1-score by 11.61%, while accuracy decreases by only 0.73 %, from 98.30% to 97.57%. Class-wise analysis further indicates substantial recall improvements for several underrepresented attack categories, although precision trade-offs remain for some classes. These findings show that restructuring extreme minority-class representations before oversampling can improve minority-oriented detection performance under severe class imbalance.
Deteksi URL Phishing yang Tangguh Menggunakan Fitur Leksikal yang Disempurnakan dan Model Pembelajaran Mesin Komparatif Rahmawati, Firnanda; Rafrastara, Fauzi Adi; Ghozi, Wildanil; Rakhmat Sani, Ramadhan; Umam, Chaerul; Alomoush, Ashraf
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.14070

Abstract

Phishing URLs continue to evade classical filtering approaches as attackers refine their obfuscation strategies, motivating machine learning-based detection methods that adapt more readily to new patterns. This work compares three classifiers, Random Forest, Extreme Gradient Boosting (XGBoost), and a hybrid CNN-LSTM model, for phishing URL classification using enhanced lexical feature engineering, syntax-aware preprocessing, and systematic component-wise ablation analysis. First, the conventional six-feature lexical representation is extended to fourteen handcrafted features by incorporating cues associated with modern phishing behavior, including suspicious keywords, top-level domain characteristics, hexadecimal encoding, and query-parameter statistics. Second, a syntax-aware preprocessing strategy selectively removes non-structural punctuation while retaining structural URL delimiters. Third, cross-source duplicate URLs are resolved through a two-stage majority-vote deduplication process to reduce redundancy and potential data leakage. Fourth, all classifiers are evaluated under a shared data preparation protocol while retaining model-specific feature representations and configurations. Experiments on the Kaggle Malicious URLs dataset achieved accuracies of 95.47%, 97.86%, and 98.59% for Random Forest, XGBoost, and CNN-LSTM, respectively, with corresponding F1-scores of 92.97%, 96.73%, and 97.89%. CNN-LSTM achieved the highest overall test-set performance with 98.59% accuracy and 99.85% ROC-AUC, while XGBoost provided the strongest performance among the tree-based models. The results indicate that combining syntax-aware preprocessing and expanded lexical features with model-specific representations provides an effective framework for phishing URL classification.
Combining Convolutional Neural Networks and XGBoost for Phishing URL Classification Bagus Nurseto, Muhammad Narendra; Rafrastara, Fauzi Adi; Ghozi, Wildanil; Rakhmat Sani, Ramadhan; Susanto, Ajib; Hamadeh, Emad
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.14097

Abstract

Cyber threat detection failures stem from the complexity of psychological exploitation tactics leveraging target anxiety and the limitations of traditional models in analyzing non-linear correlations within the structural features of malicious web URLs. To improve classification performance and computational efficiency in feature engineering, a sequential hybrid framework is developed by combining the spatial feature extraction capabilities of RGB Color Mapping image tensors via a Convolutional Neural Network (CNN) with tree-based boosting classifiers (XGBoost and LightGBM). Experiments are executed using a public dataset containing 11,055 observations, applying an 80:10:10 data splitting schema and isolated MinMaxScaler normalization to eliminate data leakage risks. The evaluation results demonstrate that the hybrid CNN + XGBoost model achieves the highest performance, with a validation accuracy of 97.83% and an F1-score of 98.06%, outperforming the replicated State-of-the-Art Random Forest baseline accuracy (97.38%). The metric stability confirms that the proposed architecture possesses robust generalization capabilities for implementation as a near real-time cyber defense system.