The proliferation of dynamic and mobile internet of things (IoT) applications, including the internet of vehicles (IoV), internet of drones (IoD), and internet of gadgets (IoG), has introduced new challenges in secure and scalable public key management. This paper proposes LHBKM, a lightweight hierarchical blockchain-based public key management framework for interconnected mobile IoT supported by mobile edge computing (MEC) adopting a three-tier hierarchy to localize registration/verification at the edge while preserving global consistency via hierarchical anchoring. Key-registration transactions are authenticated using Boneh–Lynn–Shacham (BLS) signatures, while we comparatively evaluate three hash candidates: Ascon, BLAKE3, and Keccak (SHA3). Results show that BLAKE3 provides better overall efficiency, reducing computation time by 76.47% versus Keccak while also lowering communication and storage costs by 17.60% and 15.98%, respectively. Future work will explore the integration of deep reinforcement learning to enable intelligent key caching and longterm storage optimization.

The growing digitalization of industries like automotive, transportation, urban mobility, and telecommunications is highlighting the significance of point-to-multipoint communication services. These services are essential for tasks like maximizing the efficiency of hardware and software resources and ensuring that users consistently receive software updates and communications. In order to implement precise and dependable point-to-multipoint distribution services within the network, software-defined networking is used. SDN controllers like RYU, Floodlight, and others have become the de facto standard for operating these kinds of networks. An SDN multi-controller’s efficiency greatly affects the underlying SDN infrastructure network’s adaptability and capabilities. This research makes use of a Mininet emulator to develop a multi-controller SDN architecture with a Fat-Tree topology and two RYU controllers. Throughput, latency, and roundtrip time are some of the node-to-node performance measures that will be used to determine how successful a Domain Isolation Multi-controller (DiM) network architecture is. DiM has higher throughput and round-trip time than other multi-controller algorithms. DiM is 21% faster than HyperFlow, Kandoo, and Rama at 19.99 Gbps. DiM cuts RTT by 44%, 25%, and 86% compared to HyperFlow, Kandoo, and Rama. ONOS has a slightly lower RTT (10.0 ms) than DiM (11.25 ms), but the difference is small, and DiM matches ONOS in throughput. DiM latency matches other multi-controller methods. DiM high throughput and low RTT make it ideal for scalable and responsive SDN environments.

Real-time and accurate road infrastructure monitoring is a major challenge in urban areas. Traditional methods, such as manual inspections by municipal staff or vehicular surveys using costly technologies like LiDAR or laser scanners, are prohibitively expensive, geographically constrained, and deployed infrequently. To address this, crowdsourcing has emerged as an effective approach for expanding both the coverage and frequency of infrastructure monitoring. Building on this concept, CrowPotChain introduces a novel platform that combines AI-driven pothole detection with secure blockchain-based report submission, ensuring tamper-proof and reliable crowdsourced data collection. The framework utilizes the YOLOv11s-seg model for semantic segmentation, combining convolutional neural networks (CNN) with transformer-based elements, which provides impressive detection metrics (precision: 0.889, recall: 0.894, mAP @ 0.5: 0.944). Every verified report includes geolocation, date/time, and pothole size, securely embedded in a proof-of-work (PoW) blockchain for verifiability and immutability. To examine the system’s performance, a benchmark was performed on four setups: no AI and no blockchain, AI only, blockchain only, and AI + blockchain, using batches of transactions from 10 to 100. The findings show that the no AI and no blockchain deployment provides the most rapid per-transaction time (approximately 0.010 s), followed by AI only (0.059–0.135 s), blockchain only (0.075–0.162 s), and AI + blockchain (0.145–0.696 s). Although blockchain does incur substantial overhead, its combination with AI can still serve response needs for civic infrastructure crowdsource reporting. Future development will add gamification, NFT, and IPFS to enhance participation, encourage reporting, and provide scalable decentralized storage.

As the internet and complex network infrastructures continue to expand, so does the threat of sophisticated cyberattacks, compelling organizations to adopt advanced proactive defenses. A cornerstone of these defensive strategies is the honeypot. However, existing dynamic solutions often rely on reactive deployment or centroid-based clustering (e.g., K-Means), which mathematically yields invalid, unrealistic host profiles. Because intelligent threat detection increasingly relies on high-fidelity honeypot data to analyze adversary tactics, deploying easily fingerprinted decoys fundamentally undermines downstream AI-driven defense mechanisms. To overcome this limitation, we propose DYNAMIT, an intelligent honeynet deployment system that resolves the centroid validity problem by utilizing the unsupervised K-Medoids algorithm. By combining K-Medoids with a novel hybrid Manhattan-Jaccard distance metric, DYNAMIT selects valid, existing hosts as templates based on categorical hardware and binary software similarities. The system then leverages containerization and network virtualization to simulate multiple realistic, internet-facing honeypot profiles from a single physical host, ensuring the decoys remain indistinguishable from legitimate targets. Our evaluation demonstrates that DYNAMIT accurately captures the intended number of clusters with a low relative error (18.75% for 40 hosts and 6.625% for 1000 hosts) while maintaining minimal resource overhead, establishing it as a highly scalable and robust data-generation prerequisite for modern intelligent network security.

This study investigates the potential adoption of blockchain technology within the Indonesian electricity sector to address key challenges in digital infrastructure. Blockchain technology has the potential to address the challenges by facilitating immutable and distributed storage of data across multiple network points. A two-stage methodology comprising a comprehensive literature review and selection of case studies is employed to conduct the survey. Research from reputable databases is reviewed by focusing on blockchain applications in energy systems. Key criteria such as Regulation, Implementation Readiness, Urgency, Technology Readiness Level, and Business Maturity Level are analyzed to assess deployment readiness across the main use cases in the Indonesian landscape. The review finds that five main use cases in Indonesia can be enhanced by blockchain technology, including peer-to-peer energy trading, renewable energy certificate trading, electronic billing of electricity, microgrid transactions, and electric vehicle charging transactions. Furthermore, the deployment readiness analysis suggests that electronic billing and electric vehicle charging transactions emerge as the most viable options. It is supported by conducive regulations, high urgency, and existing technological infrastructure.

The Most power plants in Indonesia still use fossil fuels to produce electricity, especially in the Java-Bali region which provides more than 70% of electricity needs [1]. Apart from that, from PLN's electricity load profile, the peak load occurs during prime time, which is when the biggest users are from the household sector. Building energy accounts for 33% of world energy consumption and 40% of world GHG emissions directly and indirectly. The target of the General National Energy Plan (RUEN) in 2025 is to achieve the target of 23% of the energy mix (renewable energy) in the electricity system. This paper proposes a deep learning method based on recurrent neural network (RNN) and some machine learning algorithm to energy modelling and energy demand forecasting for buildings. This method analyses factors including energy consumption patterns in buildings, and the non-linear relationships between these parameters on hourly, daily, weekly and monthly intervals. In this research, a building dataset was used which recorded energy usage (KWh) for 3 years from 2012-2015. The characteristics of this data set can be generalized as follows: average energy usage 1,105 kWh, median energy usage 884 kWh, mode energy usage 525 kWh, standard deviation energy usage 686.9 kWh, min energy usage 0 kWh, and max energy usage 12,372 kWh. The highest energy use trend is in the prime time range 19.00-22.00 and the lowest energy use is in the 04.00-07.00 range. The results demonstrate that the proposed method can forecast building energy demand and energy supply with a high level of accuracy, showing a RNN regressor has a MAPE value of 29.20% error range in hourly data prediction for one month ahead. This prediction model can also be applied to RE production to determine the characteristics and potential of RE in the building area. Apart from that, an optimization model can be developed to balance supply and demand for electrical energy. And it can also be used to automate the use of electrical devices in the building.

Fake news has eroded trust in credible news sources, driving the need for tools to verify the accuracy of circulating information. Fact verification addresses this issue by classifying claims as Supports (S), Refutes (R), or Not Enough Info (NEI) based on evidence. Neural Semantic Matching Networks (NSMN) is an algorithm designed for this purpose, but its reliance on BiLSTM has shown limitations, particularly overfitting. This study aims to enhance NSMN for fact verification through a structured framework comprising encoding, alignment, matching, and output layers. The proposed approach employed Siamese MaLSTM in the matching layer and introduced the Manhattan Fact Relatedness Score (MFRS) in the output layer, culminating in a novel algorithm called Deep One-Directional Neural Semantic Siamese Network (DOD–NSSN). Performance evaluation compared DOD–NSSN with NSMN and transformer-based algorithms (BERT, RoBERTa, XLM, XL-Net). Results demonstrated that DOD–NSSN achieved 91.86% accuracy and consistently outperformed other models, achieving over 95% accuracy across diverse topics, including sports, government, politics, health, and industry. The findings highlight the DOD–NSSN model’s capability to generalize effectively across various domains, providing a robust tool for automated fact verification.

Wireless bandwidth becomes increasingly limited, prompting studies to explore frequency bands higher than 100 GHz to enable multi-gigabit communication services and low-latency applications. In this context, Sub-TeraHertz communication offers vast unused spectral resources, enabling unprecedented data rates that can power advanced applications such as ultra-highdefinition video streaming, holographic communication, and massive machine-type communications. Therefore, the research aims to review the performance of millimeter wave (mmWave) and Sub-TeraHertz signal propagation in a dense environment such as an industrial complex, addressing major challenges including high propagation losses, atmospheric absorption, and signal blockage. By using NS-3 to simulate different scenarios, both inside and outside the factory, the analysis aims to characterize the channel performance and provide insight into the potential of Sub-TeraHertz communication for industrial environments. Through comprehensive simulation using the Urban Micro (UMi) scenario, the analysis models the propagation loss from a dense environment with maximum blockages. In addition to the received signal as a function of transmitter–receiver distance, the results show that the number of blockages between the transmitter and receiver, the mobility of the receiver, the frequency band, and the spectrum bandwidth substantially affect the propagation loss of signal transmission in mmWave and Sub-Terahertz Bands. These findings highlight the importance of considering indoor–outdoor disparities, blockage effects, and environment-specific network design when deploying high-frequency communication systems. The research also provides a foundation for future research to explore mitigation techniques to enhance reliability and coverage in complex industrial deployments.

Federated learning (FL) enables collaborative model training without centralizing data. However, performance can vary significantly depending on the underlying computing infrastructure. This study presents a comparative experimental evaluation of two FL deployment configurations, where the first scenario utilizes a full cloud implementation with Amazon Web Services (AWS) Lightsail, and the second scenario employs standard local virtual machines. To ensure fairness and variable isolation, both scenarios were executed using identical model settings, dataset distribution, and parameters. Training was conducted for 100 epochs, and three quantitative indicators were measured, which include aggregation time, model training loss, and classification accuracy. The results show that the first scenario achieved a lower aggregation time (0.286s) than the second scenario (1.012s), while the final accuracy remained comparable at 76.98% and 76.52%, respectively. These findings demonstrate that FL can achieve stable predictive performance across heterogeneous infrastructures, where infrastructure selection may be optimized primarily based on latency sensitivity rather than model accuracy outcomes.

The Internet of Things (IoT) has emerged as a crucial element in everyday life. The IoT environment is currently facing significant security concerns due to the numerous problems related to its architecture and supporting technology. In order to guarantee the complete security of the IoT, it is important to deal with these challenges. This study centers on employing deep learning methodologies to detect attacks. In general, this research aims to improve the performance of existing deep learning models. To mitigate data imbalances and enhance learning outcomes, the synthetic minority over-sampling technique (SMOTE) is employed. Our approach contributes to a multistage feature extraction process where autoencoders (AEs) are used initially to extract robust features from unstructured data on the model architecture’s left side. Following this, long short-term memory (LSTM) networks on the right analyze these features to recognize temporal patterns indicative of abnormal behavior. The extracted and temporally refined features are inputted into convolutional neural networks (CNNs) for final classification. This structured arrangement harnesses the distinct capabilities of each model to process and classify IoT security data effectively. Our framework is specifically designed to address various attacks, including denial of service (DoS) and Mirai attacks, which are particularly harmful to IoT systems. Unlike conventional intrusion detection systems (IDSs) that may employ a singular model or simple feature extraction methods, our multistage approach provides more comprehensive analysis and utilization of data, enhancing detection capabilities and accuracy in identifying complex cyber threats in IoT environments. This research highlights the potential benefits that can be gained by applying deep learning methods to improve the effectiveness of IDSs in IoT security. The results obtained indicate a potential improvement for enhancing security measures and mitigating emerging threats.

Indonesia’s shift toward digital land certificates introduces privacy and authenticity challenges due to the inclusion of personally identifiable information (PII) and the lack of tamper-evident verification mechanisms. These concerns have intensified following the enactment of Indonesia’s 2022 Personal Data Protection (PDP) Law, which mandates stricter controls over the publication and processing of PII. To address these issues, this study investigates six reversible string-level obfuscation methods—Blackout, Affine, Puzzle, XOR, RSA, and Scramble—to conceal PII while preserving the integrity of the land certificate. The obfuscation is implemented through a Python-based system that extracts sensitive data from fixed- coordinate zones in the PDF, applies the chosen transformation, and generates a keyfile for reversible deobfuscation. These modified certificates are then distributed using a decentralized ecosystem: obfuscated files are stored on IPFS, and their hashes are immutably logged on Ethereum via smart contracts. To assess the effectiveness of each method, we conducted performance tests with 100 iterations to measure obfuscation and deobfuscation time, memory usage, and scalability using Apache JMeter stress tests. The results reveal that lightweight methods like Puzzle and XOR offer high throughput with low resource demands, while RSA provides stronger confidentiality at the cost of higher latency and memory usage. The system sustained up to 580 concurrent users before latency degradation. This paper concludes by highlighting trade-offs between performance and security and recommends algorithm selection based on operational context and threat model.

Color blindness is a condition that affects the cone cells in the eyes, either congenital or acquired, and is classified as a moderate disability impacting a portion of the global population. This condition poses significant challenges to visual experiences, particularly in accessing digital facilities such as the metaverse. This paper explores the expansion of facilities for individuals with disabilities to access metaverse-based platforms using Oculus Quest. Dynamic RGB filters in Unity3D and Spatial.io were implemented in this study as accessibility features for metaverse-based museum environments. A key challenge in this context is ensuring accessibility for colorblind visitors who may struggle with accurate color perception. The proposed method enables colorblind users to select from various filters that modify camera colors to enhance color differentiation based on specific types of color blindness. These filters, encompassing eight types of color blindness, aim to improve visual experiences and accessibility for colorblind individuals within virtual museum environments.

Every country must attain net-zero emissions, and the way to do so is through energy efficiency. According to a report released in September 2022 by the International Energy Agency in, electrification and energy efficiency are Indonesia's top goals for reaching NZE. Currently, cooling systems (chiller plants) account for more than 50 % of building energy use. Therefore, energy efficiency in chiller plant systems offers a high potential for achieving NZE and supporting the SDGs. This study seeks to identify a new algorithm control system for a building cooling system to decrease energy consumption of the building's chiller plant. A new algorithm will be developed based on predictive model with Deep Learning Neural Network Multi Output and Multi Stack Long Short-Term Memory. The developed algorithm will next be tested by running simulations with the model of Chiller Plant. Essential parameters are discovered using a matrix correlation. Based on the matrix correlation, Condenser Water System Temperature and Wet Bulb Temperature were revealed to be the most influential parameters affecting chiller plant performance. The proposed algorithm is able to optimize Chiller Plant with the result of alleviating the use of energy by 10.72 % with less error MSE, MAE, and RMSE respectively of 0.6527, 0.8079, and 0.8079.

The accurate and timely classification of toddlers' nutritional status is critical for early intervention, particularly in remote or underserved communities with limited access to healthcare professionals. However, data security, especially for children's health data, is equally essential to ensure safe storage and access. To address these challenges, this study proposes a hybrid AI-powered chatbot that integrates ensemble learning, blockchain, and decentralized storage to support both nutritional status classification and educational interaction. The system combines a random forest model for classification with GPT-3.5 Turbo for bilingual (Indonesian–English) stunting education deployed via Telegram. Preprocessing includes standardizing, normalizing, and encoding Indonesian-language nutrition data to ensure machine learning readiness. Six ensemble algorithms are evaluated using stratified five-fold cross-validation, with classification results hashed using SHA-256 and immutably stored on the Interplanetary File System (IPFS) and a local Ethereum blockchain. The chatbot effectively manages both structured inputs and natural language queries, ensuring secure, transparent, and real-time nutritional assessments. Results demonstrate high classification performance, with the random forest model achieving the highest mean F1-score (0.9987) and the lowest deviation. Its robustness was validated by a 20% hold-out test set and stratified five-fold cross-validation, which obtained excellent balanced performance across nutritional status categories (F1-macro, precision, recall, accuracy ≈ 0.99; ROC AUC = 1.00). External validation also yielded robust and consistent results (F1-macro = 0.97, precision = 0.97, recall = 0.96, ROC AUC = 0.98, and accuracy = 0.97), demonstrating the model's generalization ability and mitigating concerns regarding overfitting. Blockchain evaluation confirmed stable and linear CID transaction throughput (blocks 29–46) with no observed latency, ensuring reliable and continuous data recording. Furthermore, gas prices decreased by ~87.5%, highlighting significant improvements in cost efficiency and scalability, which reinforces blockchain's feasibility for decentralized, AI-driven health data management. Received: 9 June 2025 | Revised: 29 September 2025 | Accepted: 31 October 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in Kaggle at https://www.kaggle.com/datasets/rendiputra/stunting-balita-detection-121k-rows and https://www.kaggle.com/datasets/jabirmuktabir/stunting-wasting-dataset. Author Contribution Statement Wa Ode Siti Nur Alam: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Riri Fitri Sari: Conceptualization, Writing – review & editing, Supervision, Funding acquisition.

Deploying private blockchain in archipelagic regions poses significant challenges due to network delays caused by geographic dispersion. This study evaluates the performance of three consensus mechanisms—Raft, Practical Byzantine Fault Tolerance (PBFT), and Proof of Authority (PoA)—to determine the most suitable for such environments. Using Network Simulator 3 (NS3), we simulate network scenarios with both fixed delay (5 ms) and varying delays (1 ms to 100 ms), alongside different node counts, block message sizes, transaction speeds, and transaction sizes. Key performance indicators such as consensus time and throughput are analyzed. The results show that Raft performs optimally under low-latency, stable conditions, whereas PBFT delivers the lowest consensus times but experiences performance degradation in environments with variable delays. PoA demonstrates the highest adaptability, excelling in highlatency conditions with consistent performance.It is also observed that uniform (same) delay performs better than varying delay across all consensus mechanisms, and that block message size significantly impacts performance more than transaction time. This research provides insights into selecting the optimal consensus protocol for private blockchain deployment in archipelagic infrastructures with diverse network conditions.

Unmanned Aerial Vehicles (UAVs) have significant roles across diverse applications, including surveillance, disaster response, search and rescue, etc. Robust communication protocols are crucial in Flying Ad-hoc Networks (FANETs) to maintain reliable connectivity between UAVs. This research examines enhancements to the Ad-hoc On-demand Distance Vector (AODV) protocol by applying the K-Means Clustering method aiming to optimize communication in FANETs. Through extensive simulations in the ns-3 environment, the proposed approach is evaluated in diverse scenarios that consider factors such as node mobility and network density. The results show that integrating AODV with K-Means Clustering improves network Quality of Service metrics, including throughput, packet delivery ratio, and end-to-end delay. This research contributes insights into the feasibility and advantages of employing clustering techniques in FANETs routing protocol.

Carbon Capture and Storage (CCS) technology emerges as a promising solution to combat CO2 emissions from industrial processes and power generation, playing a vital role in the global efforts to mitigate climate change. However, despite its potential benefits, the widespread adoption of CCS faces challenges across technical, economic, and regulatory domains. In Indonesia, a nation heavily dependent on fossil fuels, CCS presents a significant opportunity to reduce emissions while sustaining economic growth. Therefore, this paper explores Indonesia’s strategic approach to CCS initiatives, examining the regulatory framework and the current landscape of CCS projects within the country. The study discusses the contributions of Indonesian universities towards advancing CCS technology through collaborative research, partnerships, and innovative initiatives. By assessing the role of universities in addressing environmental challenges and fostering a greener future, this paper highlights the pivotal importance of universities in driving sustainable solutions for mitigating greenhouse gas emissions and achieving climate objectives. The finding concludes that Indonesian universities significantly contribute to the development and implementation of CCS through collaborative efforts and partnerships, facilitating a path towards sustainability.

The ASTM D6433-18 standard is widely used internationally in the Road Inspection System (RIS) to assess pavement distress, covering type, severity, and quantity. Automated detection of pavement distress types using vision-based methods follows these standards and requires a dataset with 19 types of pavement distress. The Road Damage Dataset (RDD) 2018 is a publicly available collection comprising 9,053 images captured on Japanese roads, each annotated with eight distinct types of road damage. However, only four of these classifications align with the categories specified in ASTM standards, namely alligator cracking, joint-reflection cracking, longitudinal and transverse cracking, and potholes. This article aims to assess the viability of utilizing the RDD dataset for Road Inspection System (RIS) purposes in accordance with the ASTM D6433-18 standard. The methodology involves the automated re-annotation of the dataset utilizing YOLOv8 models known as pseudo-labeling, followed by an evaluation to ascertain its compatibility with RIS requirements. The results suggest that the RDD-18 dataset is not suitable for conducting RIS in adherence to the ASTM D6433-18 standard. The evaluation results demonstrate less-than-optimal accuracy, which is attributed to an imbalanced distribution of instances among classes and a requirement for improved image quality. Finally, it is highlighted that the RDD 2018 dataset lacks images representing 15 additional types of pavement distress crucial for RIS applications based on ASTM standards.