Articles

Access the latest knowledge in applied science, electrical engineering, computer science and information technology, education, and health.

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

31,456 Article Results

DDoS attack detection on portmap using a bidirectional recurrent neural network method

10.11591/ijict.v15i4.pp1601-1613
Wisnu Adi Putra , Deris Stiawan , Qurhanul Rizqie , Nurul Afifah , Septiani Kusuma Ningrum , Rahmat Budiarto
Portmap is a protocol used in Unix and Linux operating systems to map port numbers to their associated program numbers. In a distributed denial of service (DDoS) attack on portmap, attackers will exploit a weakness in this protocol to send many malicious or manipulative server-running portmap requests Due to the connectionless and unauthenticated nature of UDP, the portmap server will attempt to respond to every request it receives. However, attackers often disguise this so that their IP address is hidden or spoofed. As a result, the server will attempt to reply to requests to a spoofed or invalid IP address, using server resources and causing a denial of service for valid users. A DDoS attack on portmap can be quite damaging. Services associated with portmap may become unresponsive, causing significant downtime for applications or services that rely on the protocol. This study uses the CSE-CIC-IDS 2019 dataset, especially the DDoS attack dataset on portmap. This paper also employs the correlation-based feature selection (CSF) approach to extract significant characteristics from the dataset Bidirectional RNN is used to detect DDoS assaults. Utilizing 30% testing data and 70% training data, the Bidirectional RNN approach can effectively identify DDoS attacks on portmap. Accuracy is 99.9917%, recall is 99.9917%, specificity is 99.9942%, precision is 99.994%, F1-Score is 99.88%, balanced accuracy (BACC) is 99.92%, and matthews correlation coefficient (MCC) is 99.98%. By merging the CFS feature selection technique with the Bidirectional RNN architecture, this work mainly contributes to creating a more accurate and efficient DDoS attack detection method. With this method, the detection process can be carried out faster and more precise, reducing the computational burden without sacrificing detection performance.
Volume: 15
Issue: 4
Page: 1601-1613
Publish at: 2026-12-01

Developing a machine learning-based algorithm for secure routing in autonomous vehicles

10.11591/ijict.v15i4.pp1746-1758
Shwetansh Priyadarshi , Alok Srivastava
Over the last several years, we have noticed that the administration and car companies have been refraining from investing capital in connected autonomous cars. A chain of interconnected sensors, actuators, and devices—combining lidar, GPS, radar, in-system machines, cameras, and more—allows autonomous cars to learn how to operate in real-time scenarios. However, data from recent cyberattacks and mishaps suggests that malicious attackers can access this sensor data or even the main control system through hacking, leading to a variety of losses. As time passes by, people will be less interested in connected autonomous vehicles (CAVs), and various surveys say that it will be mainly because of the absence of a proper security infrastructure. A study of cyberattacks on CAVs and strategies to eliminate them is discussed in this study. In this study, we have proposed a hybrid model solution that uses convolutional neural network (CNN), long short-term memory (LSTM), and gated recurrent unit (GRU), and we are optimizing the Hyperparameters using the dwarf mongoose optimization algorithm (DMOA). The proposed model is also tested in Comparison with other existing models, and we are able to detect DoS, fuzzy, spoofing the driving gear, and spoofing the RPM gauge attacks with an accuracy of more than 98%.
Volume: 15
Issue: 4
Page: 1746-1758
Publish at: 2026-12-01

Energy aware clustering and routing for wireless sensor networks using metaheuristics optimization algorithms

10.11591/ijict.v15i4.pp1736-1745
Sagar Mekala , Nagaraju Karinagappa , Mohan Dholvan
Wireless sensor networks (WSNs) and their applications reduce energy consumption to enhance the lifetime of the network. WSN applications require energy-efficient communication, but the network includes resource constraint tiny sensors across monitoring area. Data processing, aggregation, and clustering are energy-intensive, and deployment and discovery of nodes are complex. In order to overcome certain issues in WSN application, optimal selection of fine cluster heads (FCHs) and cluster heads (CHs) and optimal routes to the base station (BS) are proposed, still lot of optimizations are essential to enhance lifetime of the network. This research paper employs bacterial foraging and ant colony optimization (BFACO) to identify optimal gateways (FCHs), CHs, and energy-efficient paths from sensor modes to BS while optimizing node energy consumption. The algorithm dynamically creates routing paths with lower network overhead, enabling long-term network utilization and enhanced lifetime. Performance is evaluated against parallel research algorithms using metrics such as energy-usage, network lifespan, and packet delivery ratio (PDR). An energy-usage (52.8% of energy reduction over best case), packet delivery ratio (8.7% of enhancement over best case), and network lifetime (4.3% of rounds over best case) are WSN performance outcomes for 300 sensors. The BFACO using a clustering and routing algorithm outperforms over parallel research.
Volume: 15
Issue: 4
Page: 1736-1745
Publish at: 2026-12-01

Adaptive weighted multi-criteria decision making using modified dijkstra’s algorithm: integrating real-time traffic conditions with hybrid WSM-WPM for dynamic route optimization

10.11591/ijict.v15i4.pp1929-1943
Basorudin Basorudin , Handaru Jati , Nurkhamid Nurkhamid , Antoni Pribadi , Satria Riki Mustafa , Puput Dani Prasetyo Adi
Traffic congestion is a problem that needs to be solved. A smart route optimization system is needed to overcome this problem, not only using traditional methods but also using algorithms that have the ability to find the shortest route. This study proposes a framework system for multi-criteria decision making that adapts and improves the Dijkstra and weighted product method (WPM) algorithms. The Dijkstra algorithm is modified to be more integrated in solving the fastest travel route, anticipating traffic congestion, improving road safety, saving fuel, and providing real-time data. The weighted sum method (WSM)-WPM hybrid framework has the advantage of WSM additive normalization with the zero-dimensional properties of WPM to overcome different measurement units. Experimental results show a significant reduction in travel time and fuel consumption, as well as an increase in route safety percentage. This system maintains computational efficiency, and this research provides a practical and scalable solution for intelligent transportation systems (ITS), improving the driver experience and overall traffic network safety. The results of the congestion avoidance configuration with the highest hybrid score show a result of 0.4573, yielding a viable alternative route, namely (V1→V3→V6→V9→V12→V14).
Volume: 15
Issue: 4
Page: 1929-1943
Publish at: 2026-12-01

FusionGait-transformer: multimodal transformer-based framework for robust gait recognition and gait-condition classification

10.11591/ijict.v15i4.pp1902-1920
Kiran Kumar S. V. N. Madupu , Jagadeeshwar Madiraju , Naren Kodali , Soma Sekhar Goviraboyina
Gait analysis benefits from both full-body motion context and accurate temporal dynamics. Vision-only approaches degrade under occlusion and viewpoint change, while inertial-only approaches miss full body context. FusionGait transformer is a multimodal framework that fuses pose and silhouette representations with inertial time series using transformer self-attention and bidirectional cross-attention. A per-modality reliability gate weights each stream by its instantaneous signal quality, so degradation in one modality shifts the fused representation toward the other. Masked multimodal pretraining and domain adversarial training improve stability under domain shift. Evaluation uses the CMU motion capture dataset under a subject-independent protocol, with inertial signals generated analytically from the same trajectories. The framework reaches Rank-1 identification accuracy of 98.9% with equal error rate (EER) of 1.8%, and gait-condition classification F1-score of 95.7% with area under the ROC curve (AUC-ROC) of 97.6%. Reliability gating is the largest single contributor to these figures. Deployment is characterized on workstation, edge, and gateway hardware, and transmitting embeddings rather than video reduces the uplink requirement by more than four orders of magnitude, establishing feasibility for wearable healthcare settings. Reported results describe performance under controlled benchmark conditions. Clinical diagnostic claims remain conditional on validation with real wearable sensors and clinically labeled cohorts.
Volume: 15
Issue: 4
Page: 1902-1920
Publish at: 2026-12-01

Approximate computing circuits for error-tolerant applications with low energy consumption

10.11591/ijict.v15i4.pp1624-1634
Anusha Gorantla , Vamsidhar Anagani , Venkata Sridhar Gummadidala , Ramesh Babu Geddam
Energy efficiency is a crucial factor in very large-scale integrated (VLSI) circuit design, and traditional methods offer limited solutions to power intensive circuits. Recently, approximate computing techniques have emerged, enhancing key design parameters like power consumption, propagation delay, and area, albeit with a compromise in accuracy. These techniques are especially suitable for high-performance computing, as they offer reduced energy consumption while compromising some accuracy, making them ideal for use in arithmetic circuits. This paper introduces enhanced versions of approximate adders (AxAs) and subtractors (AxSCs) for fundamental operations. Building upon these designs, approximate multipliers using the Wallace tree multiplier (AxWTM) and approximate dividers (AxDs) have been developed, and their performances have been compared regarding power consumed, spatial requirements, delay time, and error frequency. Many digital image and video processing applications, along with other multimedia tools, can function effectively even with slight inaccuracies in their output results. The existing and proposed designs were applied to image squaring, image multiplication, image subtraction, and image division applications, and peak signal-to-noise ratio (PSNR) was obtained. It is found that the suggested approximate designs achieved a better PSNR for all image applications, and the results were superior to those of their counterparts.
Volume: 15
Issue: 4
Page: 1624-1634
Publish at: 2026-12-01

Automatic multi-class classification of geographical indication tagged tribal art using deep learning

10.11591/ijict.v15i4.pp1523-1531
Malay S. Bhatt , Apurva A. Mehta
In the field of computer vision, deep learning has significantly improved automated image classification, yet the classification of Indian tribal artworks remains largely unexplored. These art forms exhibit substantial stylistic overlap, variations in color palettes, and complex hand-drawn patterns, making manual identification difficult and automated classification even more challenging. Existing methods for artwork recognition are limited by small datasets, domain-specific variations, and a lack of curated tribal art collections. This study addresses these gaps by developing a deep learning based framework for the multi-class classification of three culturally significant Indian tribal arts: Manjusha, Gond, and Warli. A manually curated dataset of 900 images was augmented using geometric and color based transformations to increase robustness. The model is built on the Inception-ResNet-V2 architecture using transfer learning. Experimental evaluation demonstrates high performance, achieving 98.21% testing accuracy and stable validation trends. The proposed approach provides a scalable baseline for tribal art classification, supporting digital heritage preservation and future development of retrieval or educational systems.
Volume: 15
Issue: 4
Page: 1523-1531
Publish at: 2026-12-01

Towards accurate diagnostics: deep-transfer learning in skin disease detection

10.11591/ijict.v15i4.pp1577-1588
Vaibhavi Pandya , Vruti Parikh , Mohit Rathod , Sumitra Menaria , Nitesh Sureja
Automated skin disease diagnosis remains a significant challenge due to inherent variations in skin tone, texture, lighting, and occlusions such as hair. Addressing these diagnostic difficulties is vital for developing reliable clinical decision-support systems. This research presents an automated classification framework utilizing deep learning, transfer learning, and explainable artificial intelligence (XAI). The study utilizes the ISIC 2019 dataset, comprising images across eight diagnostic categories, including Melanoma, Nevi, and Keratosis. Six state-of-the-art convolutional neural network (CNN) architectures—Xception, DenseNet169, EfficientNetV2S, MobileNetV2, InceptionResNetV2, and ResNet101V2—were fine-tuned and evaluated. The workflow integrates advanced preprocessing, comprehensive model training, and XAI-based validation to ensure transparency in diagnostic predictions. Quantitative evaluation reveals that Xception and EfficientNetV2S outperformed other architectures, achieving near-optimal performance, with Xception reaching an area under the curve (AUC) of 100% and superior scores across all primary metrics. Conversely, MobileNetV2 demonstrated the lowest comparative efficacy, with an accuracy of 62.66% and an AUC of 78.10%. The integration of high performing CNNs with XAI techniques provides an accurate, interpretable, and scalable solution for dermatological diagnostics. This approach not only improves classification precision but also enhances clinical trust by providing visual explanations for model decisions.
Volume: 15
Issue: 4
Page: 1577-1588
Publish at: 2026-12-01

Unidirectional and bidirectional LSTM for nonlinear synthetic time series forecasting

10.11591/ijict.v15i4.pp1696-1705
Nur Haizum Abd Rahman , Hani Syahida Zulkafli
Deep learning models are sensitive to architecture choice and hyperparameter tuning. This study evaluates the forecasting performance of unidirectional long short-term memory (LSTM) and bidirectional LSTM (BiLSTM) networks using synthetic nonlinear time series data. Synthetic data were used to provide a controlled experimental setting in which the nonlinear data-generating process, sample size, and hyperparameter effects could be systematically examined. Threshold autoregressive (TAR) and self exciting threshold autoregressive (SETAR) models were simulated with sample sizes, 1000 and 3000. Each series was partitioned into training and testing sets using an 80:20 split. Both networks were trained using grid search by varying the number of hidden neurons (32, 64, 128) and training epochs (20, 50, 100). Forecast accuracy was evaluated using mean squared error (MSE) and mean absolute error (MAE), averaged across 30 replications. The results show that increasing epochs and hidden neurons improves accuracy, but the improvements decline at higher settings. BiLSTM achieves lower forecasting errors than LSTM for both TAR and SETAR datasets, and the Wilcoxon test confirms that the differences are statistically significant. The main contribution of this study is a controlled empirical framework that links the architecture, sample size, and hyperparameter settings to nonlinear regime-switching forecasting performance. The findings suggest that BiLSTM is preferable when higher accuracy is required, while LSTM remains a practical alternative for smoother nonlinear series or resource-constrained applications.
Volume: 15
Issue: 4
Page: 1696-1705
Publish at: 2026-12-01

Expert system for motorcycle damage diagnosis using certainty factor and case-based reasoning methods

10.11591/ijict.v15i4.pp1829-1840
Anthony Anggrawan , Christofer Satria , Bambang Krismono Triwijoyo , Dadang Priyanto , Hasbullah Hasbullah , Victoria Cynthia Rebecca
Motorcycles are the main mode of transportation that requires maintenance due to damage. Unfortunately, damage detection on motorcycles remains traditional, despite the potential for expert systems and artificial intelligence (AI) to facilitate more straightforward diagnosis for both motorcycle users and mechanics. That is why this research aims to develop a system with an expert system approach to diagnose damage to a motorcycle. This research applies the certainty factor (CF) expert system approach and the case-based reasoning (CBR) AI method. The study found that the expert system using the CBR method outperforms the CF method in diagnosing motorcycle damage. The CF expert system method achieves an accuracy performance of 98%. In comparison, the CBR expert system method achieves an accuracy of up to 100%, or can be equated with human mechanics in diagnosing motorcycle damage. The finding implies that CBR can effectively replace human experts in detecting motorcycle damage.
Volume: 15
Issue: 4
Page: 1829-1840
Publish at: 2026-12-01

From data to decisions: ensemble deep learning framework for maize seed variety classification with explainable AI

10.11591/ijict.v15i4.pp1648-1663
Emmanuel Asante , Obed Appiah , Ezekiel Mensah Martey , Emmanuel Effah , Eric Opoku , Ernest Kwame Ampomah
Accurate and non-destructive identification of maize seed varieties is essential for maintaining seed purity, improving crop productivity, and supporting agricultural quality assurance. Although recent studies report high accuracies using advanced imaging and deep learning techniques, many rely on hyperspectral sensors or controlled acquisition environments, or offer limited interpretability, thereby restricting real-world adoption. This study investigates maize seed variety classification using low-cost RGB imaging and an ensemble deep learning framework, focusing on three widely cultivated Ghanaian varieties: Bhihilifa, Sanzal Sima, and Wang Dataa. In collaboration with Heritage Seeds Ghana, an expert-verified dataset of 17,724 maize seed images was curated under realistic acquisition conditions with an Apple iPhone 12. A baseline convolutional neural network (CNN) was developed for benchmarking, followed by transfer learning with DenseNet169, MobileNetV2, and Xception architectures. An ensemble framework integrating DenseNet169 and Xception was proposed to improve robustness and generalization. Across five-fold cross-validation, the ensemble achieved a mean accuracy of 98.89±0.17%, precision of 98.86±0.18%, recall of 98.80±0.17%, and F1-score of 98.83±0.17%. To enhance transparency, Grad-CAM and saliency maps were employed to visualize discriminative features influencing model decisions. Results demonstrate that accurate and interpretable classification of maize seed varieties is achievable using standard RGB imagery without specialized sensing equipment.
Volume: 15
Issue: 4
Page: 1648-1663
Publish at: 2026-12-01

Federated deep learning for early detection and progression modelling of diabetic nephropathy

10.11591/ijict.v15i4.pp1453-1458
Aravind Kumar Kyrika , Sumit Kumar Banchhor
One of the main causes of end-stage renal disease is diabetic nephropathy (DN). Patient privacy-sensitive data is a restriction in centralized machine-learning approaches for the healthcare domain. This research introduces a privacy-preserving federated deep-learning (Fed-DN) method for early diagnosis of DN and modelling disease progression and its impact. Fed-DN combines safe aggregation with differential privacy (DP) using Opacus and multi-modal data sources (electronic health records, fundus imaging, and histopathology). Both simulation tests with multi-modal data sets distributed across hospitals show an efficient privacy-utility trade-off, with Fed-DN achieving near-centralized performance (AUC ≈ 0.90) (Accuracy = 87%) while maintaining a privacy budget of ε ≈ 3.1. This is a good example of the trade-off achieved between model accuracy and privacy protection. The technique does not reveal patient information, which enables collaborative, scalable healthcare analytics. Furthermore, it enhances trust among patients and institutions.
Volume: 15
Issue: 4
Page: 1453-1458
Publish at: 2026-12-01

CPU-optimized regularized convolutional neural network architecture for multimedia-based breast cancer metastasis detection

10.11591/ijict.v15i4.pp1881-1892
Guidoum Amina , Bouhenni Abdelhake
The histopathological diagnosis of breast cancer metastases, a subjective and lengthy process, benefits from advances in artificial intelligence for its automation. This study proposes an automatic classification method based on a convolutional neural network (CNN) architecture with progressive regularization (Batch normalization and adaptive dropout), trained on the PatchCamelyon (PCam) dataset. Our approach integrates a data augmentation strategy specific to medical images and advanced optimization for CPU environments, allowing us to balance performance, robustness, and computational accessibility. The architecture comprises only 94,000 parameters, 223 times fewer than leading models. Multi-criteria evaluation shows competitive results: area under the curve (AUC) of 92.92%, F1-score of 82.39%, accuracy of 88.74%, and specificity of 90.61%. A bootstrap robustness analysis provides 95% confidence intervals (AUC: [89.56% 92.91%]), and an ablation study quantifies the reduction in overfitting through progressive dropout (33.6%). This research validates that a well regulated and CPU-optimized CNN architecture can achieve clinically relevant performance while remaining accessible on standard environments, thus opening up concrete prospects for the development of practical AI tools in digital pathology.
Volume: 15
Issue: 4
Page: 1881-1892
Publish at: 2026-12-01

Evaluating usability, accessibility, and persuasion in learning management systems

10.11591/ijict.v15i4.pp2011-2024
Samantha Jim'enez , Fernanda Murillo-Mu'noz , Leticia Cervantes , Norma Candolfi , Berenice Fong , Camilo Caraveo , Alan Ram'ırez-Noriega
Learning management systems (LMSs) are widely adopted in engineering education to support content delivery, foster collaboration, and develop digital competencies. However, existing platforms often prioritize functionality over user engagement and accessibility. There is limited research that holistically evaluates LMS usability, persuasive design, and accessibility in the context of technical education, particularly during high-frequency tasks such as assignment submission. This study aims to evaluate and compare the usability, persuasive elements, and accessibility of blackboard, canvas, and Google Classroom during the assignment submission process from the perspective of engineering students. An experimental, between-subjects design was employed with engineering students who each interacted with one of the three LMS platforms. We evaluated usability, persuasiveness and accessibility. Canvas was the most efficient platform, with shorter task completion times and fewer errors. However, it received lower scores in satisfaction and persuasive design compared to Google Classroom, which stood out in credibility and user guidance but showed more variability in efficiency. None of the platforms fully complied with key accessibility criteria, such as keyboard operability and proper form labeling. These findings highlight the importance of designing LMS platforms that balance us ability, persuasion, and accessibility to better support engagement and equity in engineering education.
Volume: 15
Issue: 4
Page: 2011-2024
Publish at: 2026-12-01

Development of KabuTeach: a character-based decision support system for mushroom classification

10.11591/ijict.v15i4.pp1561-1576
Vilchor G. Perdido , Melinda A. Beninsig , Daniel Marvin B. Ramel
Mushroom misidentification remains a persistent public health risk in rural communities, causing poisoning and fatalities worldwide. This study presents KabuTeach, a web-based decision support system (DSS) for classifying mushrooms as edible or poisonous. The system is trained on the UCI mushroom dataset (8,124 instances, 22 categorical morphological attributes) using the categorical Naïve Bayes algorithm with Laplace smoothing (α=1.0) through the Rubix ML framework. The system employed a hybrid interactive structure and event-driven architecture with the model view–controller pattern, enabling real-time classification, human-in-the-loop expert validation, and SMS-based result delivery for low-connectivity rural environments. Model evaluation using a 70:30 holdout method achieved 99.63% accuracy, while 5-fold cross-validation yielded a mean accuracy of 99.66% (±0.1118%). Performance metrics, including a confusion matrix, per-fold results with standard deviations, precision, recall, F1-score, and specificity, confirm model robustness, attributed to the clean and discriminative categorical features of the dataset and the use of Laplace smoothing. Usability testing with 102 respondents using the USE questionnaire achieved a grand mean of 5.93 (agree). Reliance on the North American UCI dataset limits direct Philippine applicability; future work will focus on locally sourced, expert-labeled datasets. KabuTeach is a reliable, interpretable, and user-centered DSS that bridges traditional knowledge and modern ICT for safer rural mushroom foraging.
Volume: 15
Issue: 4
Page: 1561-1576
Publish at: 2026-12-01
1 2 3
...
2097 2098
Show 2 of 2098

Discover Our Library

Embark on a journey through our expansive collection of articles and let curiosity lead your path to innovation.

Explore Now
Library 3D Ilustration