Article Per Year (5 Year)
p-Index From 2021 - 2026
1.084
P-Index
This Author published in this journals
All Journal Jurnal Teknik Informatika C.I.T. Medicom Jurnal Mandiri IT International Journal of Enterprise Modelling Journal of Defense Technology and Engineering
Bagus Hendra Saputra
Universitas 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 Economics, Econometrics & Finance Electrical & Electronics Engineering Energy Engineering Industrial & Manufacturing Engineering Library & Information Science Mathematics Mechanical Engineering Transportation

Published : 7 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 7 Documents
Search

 1 
Multimodal deep learning framework for detection and attribution of adversarial information operations on social media platforms Nick Holson M. Silalahi; Jonson Manurung; Bagus Hendra Saputra
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

Adversarial information operations on social media platforms pose critical threats to national security, with state-sponsored actors exploiting multimodal content manipulation to conduct sophisticated disinformation campaigns. Existing detection approaches focus on single-modality analysis, lacking comprehensive frameworks for simultaneous detection, attribution, and coordination identification. This research develops an integrated multimodal deep learning framework combining RoBERTa-large transformer, Vision Transformer, Graph Convolutional Networks, and bidirectional LSTM, unified through cross-modal attention fusion with multi-task learning optimization. Experimental validation utilizes eight datasets including Russian IRA tweets (3.8M posts), Fakeddit (1M submissions), TweepFake (25K accounts), FakeNewsNet (23K articles), MM-COVID (6.7K posts), CREDBANK (60M tweets), and MEMES (12K items). Results demonstrate 93.24% detection accuracy, 79.34% attribution accuracy across 15 threat actor groups, 91.67% coordination F1-score, 88.62% narrative classification accuracy, and 448ms inference latency suitable for real-time deployment. Ablation studies reveal graph neural networks provide largest performance contribution (5.82% improvement), highlighting social network analysis importance for detecting coordinated behavior. Future directions include large-scale pre-training, adversarial training, continual learning, human-AI collaboration, multilingual expansion, federated learning, and causal inference methods.
Content-Based Filtering Using TF-IDF for a Course Recommendation System for Indonesian MSME Entrepreneurs Rr Octanty Mulianingtyas; Bagus Hendra Saputra; Rohani Situmorang; Galih Prakoso Rizky A
International Journal of Enterprise Modelling Vol. 20 No. 2 (2026): May: Enterprise Modelling
Publisher : International Enterprise Integration Association

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/int.jo.emod.v20i2.172

Abstract

Micro, Small, and Medium Enterprises (MSMEs), particularly those led by female entrepreneurs, play a vital role in Indonesia’s economic development; however, digital learning platforms often lack adaptive mechanisms that align course offerings with individual learning needs. Existing platforms generally rely on manual selection or generic categorization, creating a gap in personalized recommendation support within digital entrepreneurship education. The primary objective of this study is to develop and assess a content-based course recommendation system for Femalepreneur.id that addresses this limitation. A quantitative experimental research design was adopted using user profile data collected through questionnaires and course descriptions obtained from the platform repository. The methodology integrates systematic text preprocessing, feature representation using Term Frequency–Inverse Document Frequency (TF-IDF), and similarity computation through cosine similarity to generate personalized recommendations. The experimental results indicate Mean Precision@3 at 51.5%, Mean Average Precision@3 (MAP@3) at 42.9%, and Hit Ratio@3 at 90.9%. The precision matrix demonstrates the system can recommend relevant result until three courses as the maximum value based on the ground truth.  While, Hit Ratio matrix reveals that at least the system can recommend at least one relevant topic.  These findings confirm the effectiveness of TF-IDF in modelling textual learning features and highlight the contribution of the proposed system in strengthening personalized digital entrepreneurship learning for female entrepreneurs.
Human object detection and classification system based on thermal cameras using the YOLOv11 object detection model Muhammad Irsyaad Nurrahman; Bagus Hendra Saputra; H. A. Danang Rimbawa
International Journal of Enterprise Modelling Vol. 20 No. 2 (2026): May: Enterprise Modelling
Publisher : International Enterprise Integration Association

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/int.jo.emod.v20i2.188

Abstract

Strategic institutions such as military campuses, defense research centers, and government facilities face increasingly complex security challenges, particularly in environments with low visibility and limited manual patrol capabilities. Conventional surveillance systems often perform poorly in dark environments because they depend heavily on visible light. Therefore, this research proposes a human object detection and classification system based on thermal cameras integrated with the YOLOv11 object detection model. Thermal cameras are capable of capturing heat radiation emitted by objects, enabling effective visualization under low-light and completely dark conditions. The proposed system combines thermal imaging technology with the real-time detection capability of YOLOv11 to automatically identify and classify human objects. This research employs the Research and Development (R&D) method, including dataset collection, image annotation, data augmentation, data preprocessing, model training, and system evaluation. The dataset consists of thermal images enhanced using augmentation techniques such as cropping, rotation, brightness adjustment, and blur effects to improve model robustness. Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, and Confusion Matrix analysis. Experimental results demonstrate that the proposed system achieved an average accuracy of 86.36%, with accuracy values of 85.98% under completely dark conditions and 86.75% under dim-light conditions, indicating that the model is capable of reliably detecting and classifying human objects in low-visibility environments. These findings show that the integration of thermal cameras and YOLOv11 can contribute to the development of intelligent security systems that improve surveillance efficiency while reducing dependence on manual monitoring.
Design and implementation of a web-based administrative information system for PUSHANSIBER Rizal Gian Febriantama; Bagus Hendra Saputra; Sembada Denrineksa Bimorogo; Nadiza Lediwara
Jurnal Mandiri IT Vol. 14 No. 3 (2026): Jan: Computer Science and Field
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/mandiri.v14i3.477

Abstract

The design and implementation of a web-based administrative information system for Subbagian Tata Usaha at Pusat Pertahanan Siber (Pushansiber), Ministry of Defense, Indonesia, addressed critical inefficiencies in manual Excel and paper-based processes causing data redundancy, workflow delays, and limited real-time access. Using Rapid Application Development (RAD) methodology, this research developed an integrated system with innovative features for high-security defense environments: five-tier Role-Based Access Control (RBAC), Server-Sent Events (SSE) for real-time notifications, automated leave quota validation, and competency mapping algorithms with gap analysis. This work contributes a validated framework for administrative digitalization in high-security defense environments by integrating five administrative workflows into a unified platform with RBAC, real-time SSE notifications, and comprehensive audit logging addressing the literature gap where existing studies overlooked security-sensitive requirements in government contexts. Black Box testing with 20 scenarios validated functionalities, demonstrating quantifiable improvements: leave approval time reduced 85% (3-5 days to 4 hours), personnel data retrieval improved from 15-20 minutes to 30 seconds, missed tasks decreased from 20% to below 5%, announcement delivery achieved 97%, and data entry errors reduced to below 2%. This research establishes a replicable model for defense administrative transformation, contributing empirical evidence to public sector digital transformation literature.
Design and development of the spacelog web application for inventory management and asset tracking using QR codes at the Cyber Defense Center of the Ministry of Defense Johan Adrian Sitanggang; Bagus Hendra Saputra; Ajeng Hidayati; Hondor Saragih
Jurnal Mandiri IT Vol. 14 No. 3 (2026): Jan: Computer Science and Field
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/mandiri.v14i3.481

Abstract

SpaceLog is a web-based inventory information system developed for the Cyber Defense Center of the Indonesian Ministry of Defense to address the limitations of spreadsheet-based management, which is static, non-real-time, and lacks accountability. This study proposes a novel approach by implementing a unit-centric architecture combined with Role-Based Access Control (RBAC) specifically tailored for the high-security requirements of the defense sector. The system development utilizes the Rapid Application Development (RAD) method, built upon Laravel, MySQL, and Bootstrap frameworks. Key features include unique QR Code tracking for individual assets, hierarchical location mapping, and a comprehensive audit trail. Testing results using the Black-Box method demonstrate that all functional scenarios, including item tracking and tiered access rights (Superadmin, Section Head, Staff), operate with 100% validity. Furthermore, the implementation significantly improves operational success by transforming asset management from a manual, error-prone process into a real-time, fully auditable digital ecosystem, thereby meeting the strict accountability standards of the Ministry of Defense.
Real-time human detection on FPV drones using YOLOv11 and ESP-NOW Aria Kusumah Sastradinata; Bagus Hendra Saputra; Rifky Adishatya; Gumayang Fitri Annisa; Lusy Amelia; Belinda Zhafira; Mukhamad Ayx T Zus Rizal Tofa
Jurnal Teknik Informatika C.I.T Medicom Vol 18 No 2 (2026): May: Intelligent Decision Support System (IDSS)
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/cit.Vol18.2026.1609.pp133-150

Abstract

Conventional aerial surveillance systems still rely heavily on human operators, which may lead to visual fatigue, limited monitoring coverage, and delayed responses during security patrol operations. This study proposes a real-time human detection system for FPV drone surveillance using the YOLOv11 object detection model integrated with ESP-NOW wireless communication. The proposed system incorporates temporal validation and human-in-the-loop confirmation to improve detection reliability and maintain operator control during response activation. Experimental evaluations were conducted under morning, afternoon, and evening conditions. The proposed system achieved average confidence values of 81.25%, 78.38%, and 79.88%, with detection success rates of 71.13%, 75.94%, and 78.03%, respectively. Furthermore, the ESP-NOW communication subsystem successfully transmitted activation signals with delays ranging from 7 ms to 53 ms and maintained stable communication over distances up to 300 m. The main contribution of this research lies in the integration of YOLOv11, temporal validation, human-in-the-loop confirmation, and ESP-NOW communication into a single UAV surveillance framework, enabling reliable real-time human detection while preserving human supervision in operational decision-making.
Design and implementation of a weapon storage access control system based on hand gesture recognition and face recognition on Raspberry Pi 5 Daffa Rahman; Sunarta Sunarta; Bagus Hendra Saputra
Jurnal Teknik Informatika C.I.T Medicom Vol 18 No 2 (2026): May: Intelligent Decision Support System (IDSS)
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35335/cit.Vol18.2026.1633.pp161-170

Abstract

This study presents a multimodal biometric access control system for weapon storage facilities, integrating hand gesture recognition and face recognition through a sequential fusion architecture on Raspberry Pi 5. The sequential design activates face verification only after correct gesture authentication, optimizing computational efficiency on edge hardware while establishing a dual-layer security barrier. The gesture module combines MediaPipe Hands landmark extraction with LSTM-based temporal classification, achieving near-perfect accuracy across four gesture classes. The face module employs dlib's ResNet-34 for 128-dimensional embedding comparison, with an empirically recalibrated Euclidean distance threshold of 0.34 to eliminate false acceptance risks identified during intrusion testing. Evaluation under controlled conditions yielded 0% False Reject Rate and 0% False Accept Rate across 60 trials, with reliable GPIO-controlled solenoid actuation. Results demonstrate that sequential fusion of behavioral and physiological biometrics on a single edge device provides a viable security solution for high-risk access control applications.
Co-Authors Ajeng Hidayati Aria Kusumah Sastradinata Belinda Zhafira Daffa Rahman Galih Prakoso Rizky A Gumayang Fitri Annisa H. A. Danang Rimbawa Hondor Saragih Johan Adrian Sitanggang Lusy Amelia Manurung, Jonson Muhammad Irsyaad Nurrahman Mukhamad Ayx T Zus Rizal Tofa Nadiza Lediwara Nick Holson M. Silalahi Rifky Adishatya Rizal Gian Febriantama Rohani Situmorang RR Octanty Mulianingtyas Sembada Denrineksa Bimorogo Sunarta Sunarta