Asian Journal of Research in Computer Science
Not a member yet
792 research outputs found
Sort by
Enhancing Security in Private Network Communications through Advanced Encryption Gateways: Innovations, Implementations and Performance Analysis
No full textPrivate network communications are critical for safeguarding sensitive data in sectors such as finance, healthcare, and defense. Encryption gateways serve as pivotal components in securing these networks by integrating encryption algorithms, protocol conversion, and secure tunnel establishment. This paper explores the technical architecture of modern encryption gateways, evaluates their performance across hybrid encryption models, and introduces a novel framework for dynamic algorithm selection. A healthcare case study demonstrated a 75% reduction in data breaches while achieving zero data loss during FTP-to-SFTP migration. Research findings indicate that data security significantly improves under encrypted gateways, with performance further optimized through machine learning algorithms
Automated Identification of Indian Heritage Monuments using VGG16-Based Convolutional Neural Networks
No full textMonument recognition is a challenging task in the domain of image classification. Different structure orientations play a significant part in recognizing monuments in photographs. This paper presents a novel technique for categorizing diverse monuments based on the characteristics of their photographs. The deep Convolution Neural Networks VGG16 model is utilized to extract representations. The model is trained on cropped photos of several Indian monuments, which show a wide range of geographic and cultural variety. A monument is a physical structure dedicated to a person, event, or purpose that was built or erected. The importance of this paper to finding and classifying historical monuments accurately without any issue. We used emerging technologies for this identification purpose. without any Machine Learning and Deep Learning, are improving, accelerating image identification development, and allowing computer vision to reach new heights. There is more coverage of international landmarks and monuments, necessitating the need to link a structure\u27s physical presence to its digital presence. As a result, the monument\u27s automated identification comes into action. Almost 100 percent accuracy was predicted using the VGG 16 Model on our proposed dataset
Ensemble Machine Learning-based Heart Disease Prediction with Hyper-tuning Parameters
No full textHeart disease is the most dangerous issue in the world. The health sector also faces problems with this disease in the world. A serious investigation avoids the issue in a short time. Optimal prediction generates accurate results for this disease. This type of investigation helps doctors identify diseases and cures for health. Machine learning techniques play a crucial role in this scenario. In health industry utilizes such types of engineering techniques for speedy identification and recovery. Here we use ensemble learning with hyper-tuning parameters of the dataset. Our research observes that if I use the different machine learning models individually, then the accuracy for the decision tree is 70.37%, the Random Forest tree is 79.63%, the Support Vector Machine is 75.93%, and the Logistic Regression predicts 81.48%. But ensemble models of decision tree and Random Forest tree generate an accuracy rate is 71. 29%, and the SVM and LR accuracy rate is 76.85%
MicroTrack Vision Pro: AI-Powered Small Object Recognition for Railway Safety
No full textEnsuring operational safety on electrified railways requires the accurate detection of small foreign objects, necessitating high-precision detection algorithms. EBSE-YOLO represents an advanced algorithm dedicated to advancing small goal recognition within electrified train settings. The detection accuracy of EBSE-YOLO improves and reduces the system load by utilising advanced techniques, which include ECA-net for small goal prioritisation, BiFPN-inspired cross-level feature fusion and SPD-Conv for detail extraction and the EIOU loss characteristic for dimension alignment. Testing with different YOLOv5 configurations and Ghost CNN supplement approaches enabled the suggested approach to reach outstanding performance. EBSE-YOLO reaches a mAP precision of 97% in initial monitoring, but the system integrates YOLOv5 with Ghost CNN to surpass 98% mAP levels. EBSE-YOLO contributes benefits that extend beyond basic performance indicators because it creates substantial impacts on railway safety, together with management oversight. EBSE-YOLO applies modern model designs coupled with recognition techniques to enhance tiny goal detection capabilities and establish a system for ongoing railway safety improvement innovations. The research develops foundational guidelines that upcoming software for object detection uses to enhance railway safety operations in complex environments. The model developers are optimising it continuously to maximise performance for embedded machinery and drones that need to deploy it in real-time surveillance operations for railway safety
A Novel and Effective Multi-Model-Based Default Risk Analysis and Prediction in the Business Sector
No full textThe precise evaluation of credit risk continues to be a crucial component of prudent decision-making and risk management in the banking and lending industries in the ever-changing financial landscape of today. While traditional methods of credit risk analysis, which frequently depend on isolated modelling methodologies, have proven effective, they might not be able to fully represent the complexity of today\u27s financial settings. The need for methods that can provide increased predictive power and adaptability grows as markets change and become more complicated. To address this problem, ensemble approaches have surfaced as a strong contender, offering a framework that combines the predictive power of several models into a coherent whole. This study uses a range of machine learning algorithms, including XGBoost, CatBoost, Decision Tree, Logistic Regression, KNN, and Random Forest, to explore the potential in the field of credit risk analysis. By leveraging the unique properties of many algorithms within an ensemble framework, the objective is to improve forecast accuracy while also strengthening the robustness and adaptability of default risk assessment approaches. This introduction discusses how ensemble approaches can revolutionize credit risk analysis and establish the groundwork for a full discussion of them. It also offers insights into practical implementation considerations and empirical validations
Intelligent Fraud Prevention Information Banking: A Data Governance- Centric Approach Using Behavioural Biometrics
No full textThis study investigates the integration of behavioural biometrics and data governance in enhancing fraud prevention systems within the banking sector. Behavioural biometrics refers to the analysis of unique patterns in user interactions—such as typing rhythm, mouse movement, and touchscreen gestures—to authenticate identity continuously and unobtrusively. As financial fraud becomes more sophisticated, traditional rule-based systems are proving insufficient. This research addresses that gap by evaluating the performance and ethical deployment of intelligent fraud detection systems using real-world data. Four open-source datasets were employed—IEEE-CIS Fraud Detection Dataset, Open Data Barometer, BioCatch case studies, and Stanford AI Index Reports. Statistical techniques including logistic regression, Wilcoxon signed-rank tests, and multivariate regression were used to evaluate system effectiveness and governance impact. Results show that behavioural biometric systems achieved an accuracy of 89.9% and a ROC-AUC of 0.849, indicating strong classification performance. Post-implementation data from fifteen institutions revealed an average fraud reduction of 35.5%, with statistically significant improvement (p < 0.001). Moreover, data governance maturity was found to explain 79.3% of national fraud rate variance and 86% of system performance variability. The findings highlight the critical role of ethical data governance in enabling the secure and responsible use of behavioural biometrics. Key recommendations include enforcing robust data privacy laws, investing in biometric infrastructure, and aligning AI governance frameworks across borders. This research provides empirical evidence and practical insights for banks, technology vendors, and policymakers seeking to modernize fraud prevention in a digitally complex financial environment
Blood Pressure Prediction Using Deep Learning
No full textAlthough traditional cuff-based blood pressure (BP) monitoring is sporadic and laborious, BP is essential for cardiovascular health. We review deep learning approaches for cuff-less blood pressure estimation, such as CNNs, RNNs, Transformer models, and attention processes, and provide two new PPG-to-ABP waveform synthesis methods. The first (ASBP) maps one-dimensional PPG signals into arterial blood pressure waveforms using a VGG-16 encoder-decoder, while the second (SEANet) uses causal dilated convolutions in a calibration-free framework for continuous blood pressure estimation. Using correlation coefficient (CC), mean absolute error (MAE), and mean absolute percentage error (MAPE) measures, both models are trained and assessed on the UCI cuff-less BP dataset. The results have near-normal residual distributions and satisfy AAMI/BHS clinical criteria. An organized comparison of twenty cutting-edge studies demonstrates the variety of datasets, methodological advancements, and clinical usefulness. We present future work for reliable, generalized blood pressure monitoring with wearable PPG sensors and talk about challenges, including dataset heterogeneity and real-time deployment
A Comprehensive Review of Text Generation: From NLP to Hybrid Mechanisms
No full textThe natural language processing (NLP) field is facing a significant challenge in text generation, which is considered more complex than text understanding. The rapid expansion of electronic communication between people has made research in text generation essential. Websites across different domains now aim to respond to users using natural language. This study classifies text generation based on two principles (the level of generation and the technique used). This classification offers a comprehensive view of how text generation has developed and how different methods contribute to generating coherent and contextual text. The study recognizes deep learning as the principal approach in text generation and recommends that transforming deep learning models to include self-attention mechanisms and knowledge understanding is a promising direction for future research
How does a Peer-matching Interface for Collaborative Brainstorming Influence Cognitive Flexibility, Sustained Attention, and the Quality of Idea Development
No full textThe study aims at investigating the effects of a peer-matching interface to collaborate in brainstorming on cognitive flexibility, maintenance of attention, and the quality of the ideas generation process in general. We developed an original, adaptive algorithm that pairs the participants with complementary cognitive and interaction profiles in order to maximize creative synergy in the remote collaboration setting. The research applied to 120 subjects and incorporated quantitative measures of cognitive performance with qualitative measures of ideation output. Findings demonstrate that the peer-matching interface produced a significant increase in cognitive flexibility, enhanced sustained attention and yielded a greater number of original and well-developed ideas relative to the use of traditional brainstorming formats. The academic networking sites like LinkedIn and ResearchGate played a significant role in the recruitment of participants and the interaction after the sessions but presented issues linked to the diversity of users, engagement, and collaboration in real-time. I was the lead researcher and developed the peer-matching theoretical framework, designed the experiment, coordinated participants via academic platforms, and was the primary Age analysing and interpreting the data. Among the findings is a great insight in designing smart collaboration tools in both educational and professional innovation settings
Integrating AI-Based Therapeutic Design and Cloud Cybersecurity for Rare Genetic Diseases: A Systematic Review
No full textBackground: Rare genetic diseases affect over 300 million individuals worldwide and are often marked by prolonged diagnostic delays and scarce treatment options. The integration of artificial intelligence (AI) and cloud computing presents transformative potential for accelerating therapeutic molecule design and early diagnosis. However, this progress also raises critical concerns about data privacy, ethical use, and cybersecurity in handling sensitive genomic information.
Objective: This systematic review explores the current landscape of AI-driven therapeutic molecule design for rare genetic diseases and evaluates the role and effectiveness of integrated cloud-based cybersecurity frameworks in protecting healthcare data.
Methods: A systematic search was conducted across PubMed, Scopus, IEEE Xplore, SpringerLink, and Web of Science for peer-reviewed articles published between 2015 and 2025. Search terms included “AI in rare genetic diseases,” “therapeutic molecule design,” “cloud cybersecurity,” and “genomic data protection.” The review followed PRISMA guidelines, initially identifying 1,008 articles. After screening for relevance, duplicates, and applying inclusion/exclusion criteria, 208 articles were selected for final analysis.
Results: Key AI techniques identified include deep learning for phenotype-genotype mapping, generative adversarial networks (GANs) for molecule generation, natural language processing (NLP) for mining biomedical literature, and federated learning for decentralized, privacy-preserving model training. Facial recognition tools such as Face2Gene demonstrated higher accuracy than clinical assessments in diagnosing genetic syndromes from 2D/3D images. However, these AI models remain vulnerable to adversarial attacks, model inversion, and data poisoning. Ethical concerns such as informed consent, algorithmic bias, data ownership, and compliance with privacy regulations like GDPR were frequently highlighted. The review also noted a surge in interdisciplinary publications, with over 230,000 related studies emerging in 2025 alone.
Conclusions: AI-enabled solutions hold strong potential to revolutionize the diagnosis and treatment of rare genetic diseases through precision medicine. Yet, their integration into clinical practice demands robust cloud cybersecurity frameworks employing differential privacy, homomorphic encryption, and adversarial training to ensure data integrity and patient trust. Ethical governance must guide the responsible deployment of these technologies to ensure equity, transparency, and privacy in the era of AI-driven healthcare