Article Per Year (5 Year)
p-Index From 2021 - 2026
0.61
P-Index
This Author published in this journals
All Journal Journal of Defense Technology and Engineering
Jonson Manurung
Universtias Pertahanan Republik Indonesia, Bogor, Indonesia
Aerospace Engineering Automotive Engineering Chemical Engineering, Chemistry & Bioengineering Civil Engineering, Building, Construction & Architecture Computer Science & IT Control & Systems Engineering Decision Sciences, Operations Research & Management Electrical & Electronics Engineering Energy Engineering Industrial & Manufacturing Engineering Mechanical Engineering Transportation

Published : 3 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 3 Documents
Search

 1 
Recurrent neural network for adaptive cyber attack prediction on critical defense systems Jonson Manurung; Hengki Tamando Sihotang
Journal of Defense Technology and Engineering Vol. 1 No. 1 (2025): July, Journal of Defense Technology and Engineering
Publisher : Fakultas Teknik dan Teknologi Pertahanan, Universitas Pertahanan Republik Indonesia

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

Abstract

The threat of cyber attacks against critical defense systems is becoming increasingly complex and dynamic, requiring adaptive and proactive prediction mechanisms. This study aims to develop a Recurrent Neural Network (RNN) model to predict cyber attacks on critical defense systems with high accuracy and generalization capabilities against new attacks. The CICIDS2020 dataset was used to train and test the model, with 70% of the data allocated for training, 15% for validation, and 15% for testing. The RNN architecture was optimized by selecting the number of hidden layers, the number of neurons per layer, the activation function, and the application of dropout and regularization to minimize the risk of overfitting. The model was trained using the Backpropagation Through Time (BPTT) algorithm and evaluated using accuracy, precision, recall, F1-score, and AUC metrics. The results show that RNN outperforms LSTM, Random Forest, and SVM algorithms, with an accuracy of 97.8%, precision of 96.5%, recall of 95.9%, F1-score of 96.2%, and AUC of 0.981, and is capable of detecting rare attacks. These findings confirm the effectiveness of RNN in capturing long-term temporal patterns in cyberattack data and providing adaptive predictions for new attacks. The practical implications of this research include strengthening critical defense systems through early detection and real-time mitigation of cyberattacks, as well as providing a basis for the development of reliable proactive security systems.
Enhanced cyber attack detection using optimized random forest with SMOTE-based class balancing and feature selection Jonson Manurung; Adam Mardamsyah; Baringin Sianipar
Journal of Defense Technology and Engineering Vol. 1 No. 2 (2026): January, Journal of Defense Technology and Engineering
Publisher : Fakultas Teknik dan Teknologi Pertahanan, Universitas Pertahanan Republik Indonesia

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

Abstract

The rapid expansion of interconnected enterprise networks has intensified cybersecurity threats, while traditional signature-based intrusion detection systems remain ineffective against evolving and imbalanced attack patterns, particularly zero-day and low-frequency attacks. This study aims to develop an optimized and practically deployable intrusion detection framework by leveraging a Random Forest classifier on the CIC-IDS2017 benchmark dataset, with emphasis on robust minority attack detection, computational efficiency, and interpretability for real-world security operations. The proposed method integrates comprehensive data preprocessing, Synthetic Minority Over-sampling Technique (SMOTE) for class imbalance mitigation, feature importance–driven dimensionality reduction, and exhaustive grid search–based hyperparameter optimization within a unified machine learning pipeline. Experiments conducted on 2.52 million network flow records demonstrate that the optimized model achieves 98.14% accuracy, 96.25% weighted F1-score, and 0.993 ROC-AUC, while maintaining stable performance across all attack categories, including minority classes such as Infiltration and Botnet with F1-scores exceeding 93%. Feature selection reduced dimensionality by 58.3% and training time by 63.2% without degrading performance, enhancing deployment feasibility in enterprise intrusion detection environments. Comparative analysis confirms that the proposed approach outperforms baseline Random Forest models, traditional machine learning methods, and recent deep learning approaches while requiring significantly lower computational resources. These findings indicate that a holistically optimized Random Forest framework offers a reliable, interpretable, and operationally efficient solution for real-world network security monitoring and cyber defense systems.
Digital propaganda content detection using a transformer-based model on social media platforms Jonson Manurung; Hengki Tamando Sihotang; R. Fanry Siahaan
Journal of Defense Technology and Engineering Vol. 2 No. 1 (2026): July, Journal of Defense Technology and Engineering
Publisher : Fakultas Teknik dan Teknologi Pertahanan, Universitas Pertahanan Republik Indonesia

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

Abstract

Digital propaganda on social media has emerged as a critical challenge to democratic stability and national security. Although transformer-based language models have demonstrated promising performance in text classification, their effectiveness for propaganda detection is often constrained by subtle rhetorical manipulation and severe class imbalance, leading to biased predictions toward majority classes. This study addresses these limitations by proposing a fine-tuned RoBERTa-base model integrated with a class-weighted cross-entropy loss to improve the recognition of minority propaganda instances. The model was trained and evaluated on the SemEval-2020 Task 11 sentence-level corpus containing 14,857 annotated sentences, partitioned into 11,886 training samples and 2,971 test samples using stratified sampling. RoBERTa was selected because its robust pre-training strategy and dynamic masking enable more effective contextual representation learning, while class-weighted loss mitigates the adverse effects of the 22.80% to 77.20% class imbalance by assigning weights of 2.19 and 0.65 to the propaganda and non-propaganda classes, respectively. Under identical fine-tuning settings, the proposed model was compared with BERT-base and DistilBERT-base to ensure a fair architectural evaluation. Experimental results demonstrate that RoBERTa achieved the best performance, attaining 93.47% accuracy and a macro-F1 score of 91.82%, outperforming BERT by 1.28 percentage points and DistilBERT by 2.89 percentage points in macro-F1. These findings demonstrate that combining RoBERTa with class-weighted learning provides a robust and practical approach for propaganda detection, supporting the development of automated content moderation and misinformation monitoring systems for social media platforms. Future work will investigate multilingual propaganda detection and fine-grained propaganda technique classification.
Co-Authors Adam Mardamsyah Baringin Sianipar Hengki Tamando Sihotang R. Fanry Siahaan