The Indonesian Journal of Computer Science

An Enhanced Model of the Wireless Multicarrier Communication OFDM Systems Applied on the FPGA Platform Based on Steganography system

Ali Y. Jaber — 2025-07-20

OFDM is a promising technology due to its robustness against multipath fading. Multipath fading distorts a signal propagating in free space due to destructive or constructive interference. The evolution of 5G wireless networks has necessitated the integration of high-throughput, low-latency multi-band modulation schemes, such as orthogonal frequency division multiplexing (OFDM), into real-time hardware-optimized platforms. However, challenges related to spectral efficiency, security, and the maximum-to-average power ratio (PAPR) remain, especially when these schemes are implemented on field-programmable gate arrays (FPGAs). Steganography offers increased information security. Therefore, this paper proposes an improved OFDM model for 5G that incorporates advanced data hiding techniques using steganography to embed secure image data within ORFDM subcarriers. The proposed system is implemented on an FPGA platform, leveraging high-speed pipelines and parallelism to achieve real-time performance at a minimal resource cost. Simulation and synthesis results demonstrate significant improvements in reduced PAPR, BER, and device efficiency compared to conventional OFDM applications. The FPGA platform design takes up approximately 20% of the total available space, with very low energy consumption compared to other traditional implementation methods. The results also showed an improvement in the OFDM system's performance by reducing the BER by 30%, indicating the absence of data loss and the effectiveness of the steganography technique in these systems. This results in improved architecture performance in terms of area, power, and speed. Furthermore, the proposed approach has proven its worth in terms of security and permeability.

Enhanced Security Algorithm for Detecting Distributed Denial of Services Attacks in Cloud Computing

Coster Baloyi — 2025-07-20

Cloud Computing has the benefit of offering on-demand scalable services to its customers without having to invest much on hardware infrastructure, resources and software. Most private and public sectors are moving to the Cloud. As a result, Cloud Computing has become an ideal option due to its flexibility, scalability and cost efficiency. The existence of vulnerabilities in the network systems hosting Cloud have raised an opportunity for attackers to launch attacks in Cloud Computing. The intruders attack business applications such as webservers, financial servers, and other servers exploiting Distributed Denial of Service (DDoS) attacks. This paper proposed a Real-Time Network Traffic Attack Detection (RTNTAD) algorithm to detect DDoS attacks using real-time dataset to mitigate DDoS attacks. MATLAB was employed to evaluate the performance of RTNTAD. The proposed RTNTAD algorithm has achieved 99.2% detection rate, 99.5% classification of DDoS attacks, 0.9% connectivity time out and less than 18% false positive.

Enhanced Fake News Detection with Domain-Specific Word Embeddings: A TorchText-Based Method for News Semantics Representation.

Sikhumbuzo Ngwenya — 2025-07-20

The prevalence of misinformation in digital media highlights the need for effective fake news detection methods. This paper presents a novel approach that leverages domain-specific word embeddings, trained specifically on news content, to improve the accuracy of fake news classification. Using TorchText, we generated 128-dimensional embeddings, optimized with Bi-LSTM and GRU models, achieving a test accuracy of 93.51% with a margin of error of 0.255. Two models were developed to classify fake news based on news headlines. The first model using pre-trained embeddings achieved a test accuracy of 96.51% with a margin of error of 0.102, and the second model trained without pre-trained embeddings, resulting in slightly worse resulting in a slightly lower accuracy of 96.23% with a loss of 0.104. The comparison highlights the significant impact of domain-specific integration on model performance. This study demonstrates the value of custom integration to improve semantic representation and fake news detection accuracy, providing a powerful tool to combat misinformation.

Enhancing Diabetes Prediction Accuracy Using Stacked Machine Learning and Deep Learning Models: A Public Health Approach

Md Ziarul Islam — 2025-08-01

Diabetes mellitus is a growing public health issue in Malaysia, affecting 7 million adults aged 18 and older. By 2025, 20.1% of Malaysians will have diabetes, with the International Diabetes Federation predicting 5 million by 2030. A study aims to improve diabetes prediction accuracy and reliability. The Indian PIMA Diabetes dataset was used to develop stacked machine learning and deep learning models, with 70% ML and 30% DL achieving optimal results. The weighted soft voting ensemble (70% ML, 30% DL) outperformed individual stacking models in terms of reliability and balanced performance, improving diabetes classification with 75.65% accuracy, 67.89% precision, and 81.41% ROC-AUC. The ensemble method, optimized for medical diagnosis tasks, showed improved accuracy, robustness, and generalization. However, ethical considerations, data privacy, and algorithmic biases are crucial for maximizing AI's potential in diabetes care, highlighting the need for scalable solutions.

Implementing an Information Verification System to Prevent Academic Fraud by Employees Using a Hybrid of ANN and RF Algorithms

Lebogang Vinnas Lebopa — 2025-08-03

Academic fraud, particularly the falsification of qualifications, poses a growing threat to organizational integrity and professional credibility. This study proposes an Information Verification System (IVS) to combat employee credential fraud using a hybrid of Artificial Neural Network (ANN) and Random Forest (RF) algorithms. The method follows a two-step process: first, ANN extracts key certificate features, such as digital signatures, logos, and serial numbers, then RF classifies the certificate as authentic or fraudulent based on these features. Tested on 4,830 certificates from Mopani TVET College, alongside 1500 replicas, the system achieved near-perfect results: 98.90% accuracy, 96.75% precision, 99.33% recall, and a 98.03% F1-score, outperforming Recurrent Neural Networks (RNN), Support Vector Machines (SVM), and Logistic Regression models. By integrating with institutional databases, the IVS offers a scalable, secure solution to automate verification processes so that only legitimate qualifications are accepted. These results suggest that the proposed IVS offers a scalable and secure solution for institutions and employers, significantly improving the efficiency and reliability of academic credential verification.

Unraveling the Structure of India’s Railway Network: Insights from Network Analysis

Fadil Al Afgani — 2025-08-03

India's railway station network is a vast and complex system that plays a crucial role in the country's transportation infrastructure. In this analysis, we will explore the network of Indian railway stations using network analysis techniques. Network analysis is a statistical approach to analyzing the relationships between variables in a network. A network is a graphical representation of the relationships (edges) between variables (nodes). The study will involve constructing a network representation of the railway station network, where each station is represented as a node, and the connections between stations are represented as edges. This visualization allows us to identify the type of network, communities, overlapping communities, cascade failure, and heterogeneous information network within the network. Based on the analysis results, the formed network is a scale-free network. The community detection analysis using the Leiden algorithm shows that there are 23 clusters formed with a quality value of 0.97662. Overlapping communities are present when the value of K ≤ 3, and there is the potential for cascade failure or an epidemic when the node with the highest degree is assigned the status of infected. The formed India's railway station network is a HINs (heterogeneous information network) as it consists of various types of entities with different characteristics.

RTSO: Comprehensive Framework for Real-Time Frequency Channel Occupancy and Spectrum Hole Detection

Elesa Ntuli — 2025-08-21

Efficient spectrum utilization remains a key challenge in modern wireless communications, especially in dynamic environments with limited spectrum availability. This paper introduces Real-Time Spectrum Optimization (RTSO), a framework that combines Geo-Location Spectrum Databases (GLSDBs) with real-time spectrum sensing to detect frequency channel occupancy and identify spectrum holes. RTSO uses advanced energy detection techniques, including Additive White Gaussian Noise (AWGN) modelling, to distinguish between idle and occupied channels accurately. It incorporates mathematical tools such as occupancy time and Frequency Channel Occupation (FCO) metrics for effective spectrum analysis. A notable feature is a revisit-time-based sensing mechanism that infers channel status during intermittent scans. Practical evaluations demonstrated improved detection accuracy, reduced false alarms, and better decision-making for dynamic access to available channels. Key performance metrics, including latency, bandwidth, and error rate, were compared with baseline methods, showing substantial gains in efficiency. This work provides a valuable contribution to cognitive radio systems and dynamic spectrum access, paving the way for more intelligent and adaptive spectrum management strategies in real-time communication networks.

Framework for Enhancing Interoperability, Data Exchange, and Security in Healthcare through Blockchain Technology

Vimbai Alice Muderere — 2025-08-03

The healthcare sector is changing, such as fragmentation issues, the sharing of data, and the security of protected health information. Traditional systems tend to work independently or in silos, resulting in disjointed patient records and system inefficiency. With more trusted healthcare providers, patients relying more on digital solutions than ever, the urgency for a consistent data management solution has never been greater. This systematic literature review (SLR) aims to investigate the existing framework, factors, opportunities and challenges of blockchain technology in healthcare systems. The integrative approach was done according to the PRISMA guidelines. A literature search was carried out on various electronic databases, including PubMed, IEE Xplore, and ACM Digital Library, which gave a total of 832 articles, to begin with. Based on set scale criteria, 18 studies were deemed relevant for analysis. The findings indicate that blockchain technology holds promise due to its ability to facilitate secure and easy data sharing through immutability, cryptographic methods, and the removal of centralized authorities. However, there is a challenge of interoperability, data exchange and security within the healthcare systems and other technologies. This study contributes to the body of knowledge by developing a conceptual framework that helps policymakers, researchers, and practitioners that act as guide to effectively implement blockchain technology in healthcare. The framework addresses key considerations of traditional systems, such as scalability, interoperability, security, and regulatory compliance, and offers a structured approach to resolving current challenges.

Automation and Selection Technique for Regression Testing: An Empirical Analysis

Muhammad Hilman — 2025-08-27

Software testing, particularly regression testing, is a process that is required when changes are made to the software or its environment to ensure that the software continues to perform as expected. Motivated by real industry needs, this study reports on the experience of transitioning from manual to automated regression testing in one of the mobile applications at PT. XYZ. Prior to this study, regression testing was conducted manually, resulting in significant costs and inherent subjectivity. Test automation is then applied to the activities of test execution and test result integration as an effort to increase test productivity and efficiency. This study aims to find an efficient testing alternative by separating the flow that runs tests related to changes from the flow that runs all tests. Based on the analysis of the tested application, each flow has its trade-offs. The results show that test automation can provide benefits for regression testing, application releases, and software engineering flow. The framework presented in this paper aims to serve as a guideline for other industrial applications with similar specifications that are also considering implementing test automation.

A Dynamic Framework for Optimizing Spectrum Utilization and Interference Mitigation in White Space Networks

Elesa Ntuli — 2025-08-21

This study presents a framework for optimizing spectrum utilization and reducing interference in White Space (WS) networks using the Interference Mitigation Decision Framework (IMDF). The IMDF combines Geo-Location Spectrum Databases (GLSDs), reactive spectrum sensing, and Software Defined Radios (SDRs) to address the limitations of traditional spectrum allocation methods. The IMDF enhances allocation, reduces interference, and improves network performance by monitoring real-time spectrum usage. Simulations comparing IMDF with traditional GLSD-based methods show a 70% bandwidth saving, compared to 40% in traditional approaches. Additionally, IMDF reduces interference events by 30%, improving Quality of Service (QoS) and mitigating Cross Network Interference (CNI). With dynamic spectrum management, IMDF achieves 70% spectrum utilization, while traditional systems only reach 40%. These results demonstrate IMDF's effectiveness in dynamic environments, offering a robust solution for wireless service demand and interference mitigation in increasingly WS networks. The IMDF’s adaptability, combined with its efficient resource management, makes it a promising framework for the future of spectrum allocation in increasingly congested network environments.

Optimization of Corn Crop Nitrogen Percentage Using Genetic Algorithm

Aidil Adrianda A — 2025-08-28

Corn is one of the strategic commodities in food fulfillment in Indonesia. Despite being one of the strategic commodities for food security, corn production is still far from meeting total consumption. One of the main factors for increasing yield is the availability of nutrients, especially nitrogen. This research aims to determine the optimal nitrogen percentage to maximise corn production using genetic algorithm. Simulations were conducted using the genetic algorithm method with parameters such as population size, maximum number of generations, mutation rate, as well as Bayesian approach for the crossover method and a Gaussian distribution for mutation. The results showed that the more generations used, the better the accuracy of the curve approach to the actual data, with an optimal nitrogen value of 1.506% in the 500th generation and a production yield of 227,718 bu/ac or 15.325 ton/ha.

Comparative Security and Performance Evaluation of IPFS and Filecoin for Off-chain Blockchain Storage

Godwin Mandinyenya — 2025-08-28

The increasing demand for secure, scalable, and decentralized data management in blockchain ecosystems has intensified the need for eefective off-chain storage solutions. Traditional blockchain infrastructures offer limited storage capacity, prompting the integration of decentralized protocols such as the InterPlanetary File System (IPFS) and Filecoin. While both enable distributed data sharing, they differ significantly in architecture, incentive mechanisms, and security assurances. This study presents a systematic literature review (SLR) of 35 peer-reviewed studies, combined with a technical evaluation of IPFS and Filecoin across five critical dimensions: performance, security, incentive models, integration feasibility, and application-specific suitability. Empirical findings indicate that IPFS provides faster data retrieval (average latency ~210 ms) and simpler integration, making it well-suited for low-risk, real-time data scenarios. However, it lacks native incentivization for long-term data persistence. In contrast, Filecoin offers higher data availability (~99.9%) and verifiable storage proofs via its token-based reward system, enhancing durability and auditability, albeit with increased latency and operational overhead. The analysis reveals that neither protocol alone fully addresses the security–scalability–persistence trade-off inherent in decentralized systems. Instead, the results advocate for hybrid architectures that combine IPFS’s performance strengths with Filecoin’s robust data assurance features. This paper contributes a structured decision-making framework to support the selection and deployment of context-appropriate off-chain storage models. The findings aim to guide researchers and practitioners in designing resilient, privacy-preserving blockchain infrastructures, particularly in domains where data integrity, verifiability, and long-term accessibility are essential.

Machine Learning and Transformer-based Model for Sentiment Analysis of Indonesian E-Commerce Reviews

Wahyu Widyananda — 2025-08-30

The growth of e-commerce in Indonesia has produced a large volume of user-generated reviews, which contain valuable knowledge for business decisions. However, analyzing this unstructured text data manually is inefficient. The purpose of this study is to improve the performance of sentiment classification on Indonesian e-commerce reviews using machine learning and transformer-based models. The test method is carried out using a public e-commerce review dataset. Three models are evaluated: Multinomial Naïve Bayes, Support Vector Machine (SVM), and IndoBERT. For machine learning models, text pre-processing is performed, and features are extracted using TF-IDF. For the transformer-based model, a fine-tuning approach is used. The results show that the IndoBERT model produces better classification accuracy than the other tested models. For the given dataset, this method obtains 94,1% in accuracy, outperforming both SVM (89,5%) and Multinomial Naïve Bayes (84,2%). The IndoBERT model, despite its higher computational cost, is the most effective for this classification task.

Navigating the Frontier: Responsible AI in Practice: Governance, Applications, and Future Directions

Naresh Tiwari — 2025-06-16

Artificial intelligence systems are increasingly deployed in consequential domains, raising critical questions about governance, domain-specific applications, and emerging challenges. This paper examines the evolving landscape of responsible AI implementation across regulatory frameworks, high-stakes domains, and future research directions. It analyzes diverse regional governance approaches—from the EU's comprehensive risk-based regulation to the US's sectoral framework and East Asian models—alongside industry self-regulation mechanisms including standards, certification programs, and auditing methodologies. The research investigates domain-specific responsible AI practices in healthcare, criminal justice, financial services, and education, identifying tailored approaches to fairness, transparency, privacy, and stakeholder engagement. The paper further explores emerging challenges including foundation model governance, environmental sustainability, global equity, and AI systems reasoning about ethics. It concludes by mapping promising interdisciplinary research directions, addressing persistent knowledge gaps, and identifying essential methodological innovations and infrastructure needed to advance responsible AI practice. This comprehensive analysis offers researchers, practitioners, and policymakers practical frameworks for implementing responsible AI in an era of rapidly expanding capabilities.

Enhanced Detection of IoT-Based DoS Attacks Using A Hybrid ANN-RF Classification Model

Solomon Bulelani Ndaba — 2025-08-30

Denial of Service (DoS) attacks pose a significant threat to the integrity and availability of Internet of Things (IoT) networks, where interconnected devices are increasingly targeted due to their vulnerabilities. These attacks overwhelm systems with excessive traffic, disrupting legitimate services and potentially compromising sensitive data. Traditional detection methods often rely on predefined signatures, which struggle to keep pace with the evolving tactics employed by attackers. This study introduces a novel hybrid detection algorithm that integrates Artificial Neural Networks (ANN) and Random Forest (RF) classifiers, termed ANN-RF, to enhance the detection of DoS attacks in IoT environments. The ANN-RF model was evaluated based on critical performance metrics, including detection accuracy, False Positive Rate (FPR), and latency. Experimental results obtained through MATLAB demonstrate that the ANN-RF model achieves a detection accuracy of 93% and a low FPR of 5% when detecting 30 attacks, significantly outperforming standalone ANN and RF models, which recorded accuracies of 82% and 87%, and FPRs of 15% and 10%, respectively. Additionally, the ANN-RF model consistently maintains high detection accuracy, reducing false alarms and enhancing reliability as the number of attacks increases. Thus, the proposed ANN-RF model has strong potential to enhance real-time security in IoT networks by offering a scalable, accurate, and adaptive solution for DoS attack detection, with practical applications across domains such as smart homes, healthcare, and industrial control systems.

A Review of Vulnerability Detection Algorithms in Software Code

Zelda P. Ramahlo — 2025-08-30

Detecting software vulnerabilities is essential to keeping modern systems safe in the face of increasingly sophisticated cyber threats. This paper offers a clear and accessible overview of how vulnerabilities are currently identified, reviewing traditional, machine learning (ML), and hybrid approaches. Traditional techniques such as static and dynamic analysis are still widely used but often suffer from high false positive rates and struggle to adapt to new and evolving threats. In contrast, recent ML developments, especially those involving Random Forest (RF) and Convolutional Neural Networks (CNN), have shown significant promise in improving detection accuracy, feature extraction, and classification. Decision Tree methods remain valued for their transparency, while CNNs and other deep learning tools excel at recognizing structural and spatial patterns in code. Combining these strengths in hybrid models integrating effective feature selection with powerful pattern recognition has the potential to deliver more accurate results and reduce false alarms. However, persistent challenges remain, including limited dataset diversity, weak resilience against adversarial attacks, and the need for real-time adaptability. By bringing together the latest research and practical insights, this review aims to guide developers, security analysts, and organizations in creating more robust, automated, and adaptive security tools capable of meeting the fast-changing demands of software vulnerability management.

Evaluation of Selected Base Models for Technostress Detection

Sunday Oladipo — 2025-08-22

The widespread use of technology has led to an increase in technostress which is a phenomenon where individuals experience stress and anxiety due to their interactions with technology. As social media platforms become increasingly integral to daily life, detecting technostress from online interactions has become a pressing concern and an avenue to enrich the research in the area of detecting technostress. This study evaluates the performance of selected base models on X (Twitter data). Also, the study investigated the effectiveness of a feature extraction technique for the improvement of the model performance through data preprocessing. The study made use of the dataset of X posts (Sentiment140) obtained from the Standford University. The extracted features were used to train and evaluate four base models: Random Forest (RF), Extreme Gradient Boosting (XGB), Gradient Boosting (GB), and Light Gradient Boosting Machine (LGBM). The performance of each model was evaluated based on accuracy, precision, recall, F1-score and Kappa statistics. The RF model outperformed other base models with accuracy, precision, recall, f1-score, and Kappa score values of 88.03%, 85.98%, 85.68%, 85.79% and 79.81% respectively. The results highlight the importance of preprocessing and feature extraction techniques in improving model performance; contributes to the development of more effective technostress detection systems and provide insights into the application of machine learning algorithms for analyzing online interactions.

Development of Security Operations Center Capabilities in a Public Cloud Environment: A Case Study of PT. XYZ

M Ryan Fadholi — 2025-08-03

Cybersecurity has become a major challenge for financial institutions, including PT. XYZ. Although PT. XYZ has established a Security Operations Center (SOC) to safeguard its digital assets, the current SOC team lacks optimal capability to monitor the organization's newly adopted public cloud environment. This gap increases the risk of undetected cyberattacks targeting the cloud infrastructure. This study aims to develop recommendations for enhancing SOC capabilities in PT. XYZ’s public cloud environment using the Design Science Research (DSR) method. The initial SOC condition was analyzed through document review and observation. Capability gaps were identified through focus group discussions (FGD) guided by the SOC-CMM screening tool. The NIST Cybersecurity Framework (CSF) was then employed as the foundation for defining target capabilities. The study resulted in a set of 35 practical recommendations to improve the SOC team's capabilities, categorized according to the SOC-CMM domains.

Dogfight from the perspective of game theory

Rafi Prayoga Dhenanta — 2025-08-03

Dogfight is one of many scenarios happening in a battle for air-superiority. This research delves deeper into dogfight, using the perspectives of game theory. The purpose of this research is to model a strategy that can be used in a dogfight. This research models dogfight into game theory’s extensive-form-games and then simulates the model ccompuationally. From the simulation, the model developed in this research increases the winning rate of a certain player significantly.

Load Classification Using Feed-Forward Neural Network in Multi-Storey Buildings

Ony Armanto — 2025-08-03

Electricity consumption continues to increase year by year, leading to inefficiencies in energy management. This issue has become a major concern in modern power systems, particularly in energy monitoring systems based on Smart Grid technology. As the use of technology becomes more accessible, energy loads also grow significantly. Therefore, the ability to identify the types of electrical loads used in an installation is crucial, necessitating the implementation of load classification systems. To support the performance of electrical load classification, a Feed-Forward Neural Network (FFNN) is utilized. The results of this study show that the classification model achieved an accuracy of 99.03% with an error rate of 6.43%, and the RSME 0.098, indicating excellent classification performance

The Digital Transformation Towards Responsive Website Using Content Management System

Maryam — 2025-08-29

In the digital era, organizations are expected to follow with technological advancements, including the need to provide informative and engaging users.. This study aims to describe the transformation of a da'wah organization in delivering information as part of a broader digital outreach strategy. The main focus to utilize the WordPress CMS as a solution to enhance website responsiveness and user reach. A practical approach was adopted through the Web Development Life Cycle method, starting from analysis requirements to technological implementation. Key aspects considered include responsive design, content integration, and user experience. Functionality testing using the black-box method confirmed that the system performs as expected. The results demonstrate improved content accessibility, optimal website responsiveness across devices, and increased user engagement. This system can serve as a model for other organizations seeking to adopt similar technologies to expand the da'wah.