Fraud Detection Research Papers - Academia.edu (original) (raw)

2025, Green Inflation: International Journal of Management and Strategic Business Leadership

The objective of this study is to empirically assess the tactics that the CBN of Nigeria has implemented as the key component of the new naira note policy. The research employed a survey design with a sample size of about fifty (50) responders, comprising accountants and auditors from ten (10) ministries chosen from the Federal Capital Territory (FCT) of Abuja, Nigeria. At the 5% significant level, the hypotheses are tested using analysis of variance (ANOVA). The study's conclusions showed a strong association between forensic accounting and the litigation support service provided by Nigerian courts, as well as the effectiveness of forensic accounting in the country's public sector in identifying fraud. In a similar vein, the CBN of Nigeria uses techniques and policies that work well to stop fraud. Therefore, the study suggests that the public sector implement a continuous improvement in the internal control officers responsible for their acts. It should be mandatory for forensic accountants to receive the necessary training in forensic accounting techniques. In order to lower the prevalence of fraudulent activities in Nigeria, public sector employees should also embrace honesty, accountability, fairness, and objectivity as fundamental moral obligations. This study contributes to the existing literature by highlighting the critical role of forensic accounting in enhancing transparency and accountability within the Nigerian public sector, ultimately fostering a more robust financial environment.

2025, Deep Science Publishing

As FinTechs have been leveraging technology to provide more efficient ways to deliver banking services to their customers, traditional banks have been delegating their risk infrastructure responsibilities to others or building technological capabilities to innovate their services. While innovative technology aims to lower the cost to develop a base product, a novelty is expected in personalized service delivery or decision-making, availability, and multi-functionality in a single applet. The use of Artificial Intelligence tools to facilitate those techno-commercial characteristics is becoming more and more widespread. However, as a critical piece of the technology stack made available and accessible as a Service, AI in its various forms is also contributing to a shadow banking ecosystem of banks and financial institutions creating challenges for the conduct of monetary policy and the regulation of economy-wide financial stability risks. This has led banks to explore the use of AI to be innovative in their service invention delivery

2025, Deep Science Publishing

2025, Journal IJETRM

Online payment fraud has become a critical challenge in the digital economy, leading to substantial financial losses
and eroding consumer trust. The rise of web surfing and online shopping, so came the use of credit cards for online
transactions, as did the prevalence of online financial fraud. This study focuses on developing a machine learningbased system to detect and prevent fraudulent transactions in online payment platforms. The proposed solution
involves data preprocessing, feature engineering, and the selection of appropriate machine learning models such as
Logistic Regression, XG Boost Classifier, Random Forests, and SVC. Given the imbalanced nature of the dataset,
where fraudulent transactions are rare, advanced techniques are employed to enhance model accuracy. The
evaluation metrics include accuracy, confusion matrix. The system is designed for real-time deployment, offering a
robust mechanism to reduce fraudulent activities and improve the security and reliability of online payment systems.

2025, IJETRM

This study investigates the effect of the damping coefficient on the transverse displacement and rotation of a prestressed shear beam resting on an elastic foundation and subjected to a moving load at a constant velocity. The governing equations are coupled second-order partial differential equations. To facilitate the analysis, the finite Fourier series method was utilized, converting these equations into a set of coupled second-order ordinary differential equations. The resulting equations that characterize the motion of the beam-load system were then solved using Laplace transformation alongside convolution theory to derive the solutions. Analyses were performed to assess the effect of the damping coefficient on both the transverse displacement and rotation of prestressed shear beams of different lengths when subjected the moving load at different velocities respectively. Furthermore, the research investigates the effect of the damping coefficient on the critical velocities of the vibrating system. The results indicate that the transverse displacement and rotation of the beam significantly decrease as the damping coefficient increases. Additionally, it was observed that an increase in the damping coefficient corresponds to increase in the critical velocity, suggesting a more stable dynamic system. Practically, this connotes the significance of the damping coefficient in enhancing the dynamic stability of the beam under the influence of a moving load.

2025, IJETRM

This study examines the awareness, perception, and participation of women in Tamil Nadu with respect to life insurance. Despite the increasing financial literacy and empowerment of women, their involvement in financial instruments such as life insurance remains comparatively low. The study aims to assess the level of awareness, explore attitudes toward insurance, and identify barriers to participation. A combination of qualitative and quantitative data was collected from various districts across Tamil Nadu to provide a comprehensive understanding of the issue. The findings suggest that while basic awareness of life insurance exists, inclusion remains limited due to socio-cultural, economic, and informational barriers. The study highlights the need for targeted financial literacy programs and inclusive policy initiatives to improve women's participation in life insurance schemes.

2025, IJETRM

With the increasing realism of AI-generated images, distinguishing them from real ones has become a growing challenge. Previous studies have used Convolutional Neural Networks (CNNs) for classifying synthetic and real images, but CNNs present several limitations. They often fail to focus on critical regions, depend heavily on increasing depth-raising computational costs-and lack global context understanding due to localized receptive fields. To overcome these issues, this work introduces a hybrid deep learning approach combining three advanced architectures: Attention-CNN with CBAM, EfficientNetB0, and Vision Transformer (ViT). CBAM enhances CNNs with spatial and channel attention, improving focus on key image features like textures and artifacts. EfficientNetB0 applies compound scaling to optimize network depth, width, and resolution for better performance with fewer resources. ViT captures global dependencies by treating images as patch sequences, enabling recognition of subtle, long-range patterns. The proposed ensemble was trained and evaluated on the CIFAKE dataset, which contains both real and AI-generated images. It achieved an accuracy of 96.81%, significantly outperforming standard CNNs. This study highlights the advantages of combining attention mechanisms and transformer-based models for more accurate and efficient synthetic image detection.

2025, Journal IJETRM

Federated Learning (FL) is crucial in situations where centralization is restricted by privacy regulations or concerns because it allows collaborative model training across distributed clients while maintaining data privacy. Multi-Layer Perceptrons (MLPs), which are commonly used in traditional FL implementations, have trouble with complicated feature relationships and complex pattern recognition tasks in image classification domains. We propose Fed-KAN, integrating Kolmogorov-Arnold Networks into federated learning to address conventional architectural limitations. Our approach leverages KANs' superior functional approximation capabilities to enhance distributed image classification. We developed two specialized variants: EfficientKAN, optimizing communication through parameter sparsity, and FastKAN, accelerating convergence through adaptive learning rates. These innovations directly target non-IID (non-independent and identically distributed) data distribution challenges and communication overhead inherent in federated environments. Our comprehensive tests on FashionMNIST show that Fed-KAN performs noticeably better than traditional Fed-MLP techniques, with EfficientKAN attaining 94.2% test accuracy as opposed to Fed-MLP's 82.6%. Even our FastKAN variant outperformed traditional techniques, achieving 90.1% accuracy. Remarkably stable, EfficientKAN consistently performs well throughout 100 communication rounds. The technique works especially well for distinguishing between fashion items that have similar visuals. Fed-KAN is a valuable innovation for privacy-preserving distributed learning in difficult image recognition tasks, as demonstrated by these measurable improvements (11.6% accuracy gain with EfficientKAN).

2025, Journal IJETRM

Liver cancer remains one of the most lethal malignancies globally due to late diagnosis and limited treatment
options. Magnetic Resonance Imaging (MRI) has emerged as a powerful non-invasive diagnostic tool, offering
superior soft tissue contrast and functional imaging capabilities for liver tumor detection and characterization.
This study proposes a unique approach that leverages MRI scans integrated with artificial intelligence (AI) based
algorithms to enhance the early diagnosis and accurate classification of liver cancer. The methodology involves
pre-processing MRI images to reduce noise, followed by segmentation using advanced deep learning models like
U-Net and ResNet. These models are trained on annotated datasets to identify and differentiate benign from
malignant lesions with high precision. Furthermore, radiomic features extracted from segmented images are fed
into machine learning classifiers such as Random Forest and Support Vector Machine (SVM) to improve
diagnostic accuracy. The proposed system demonstrates exceptional performance in identifying liver tumors,
outperforming traditional diagnostic methods in sensitivity, specificity, and overall accuracy. By combining
imaging data with AI, the framework not only supports radiologists in clinical decision-making but also reduces
inter-observer variability. This novel integration aims to revolutionize liver cancer detection, enabling early
intervention and improved patient outcomes. Future directions include expanding the dataset diversity,
incorporating multimodal imaging, and real-time deployment in clinical settings. The findings of this research
highlight the transformative potential of AI-enhanced MRI analysis in liver oncology, pushing the boundaries of
personalized medicine and precision diagnostics.

2025, Deep Science Publishing

2025, INTERNATIONAL JOURNAL OF RESEARCH AND ANALYTICAL REVIEWS

The use of machine learning algorithms for predictive maintenance in manufacturing is investigated in this work. It uses secondary data from published studies, publications, and online journals and employs a desktop research methodology. The results show that there is a substantial methodological and contextual gap in the use of machine learning approaches to predictive maintenance. When it comes to anticipating equipment problems and optimizing maintenance schedules, sophisticated algorithms like deep learning and ensemble approaches show exceptional accuracy. High-quality data and real-time monitoring are important facilitators of these developments, but obstacles like the need for computational resources and the complexity of implementation continue to be major obstacles. By using the Theory of Predictive Analytics, Machine Learning Classification, and Anomaly Detection to frame the analysis, the study advances theory. Adopting complex algorithms, making investments in reliable data collection methods, and resolving implementation issues are examples of practical suggestions. The study emphasizes for policymakers the significance of creating frameworks that facilitate adoption while reducing cybersecurity and data privacy threats. To further improve predictive maintenance results and progress the manufacturing industry, further study is encouraged to investigate hybrid methodologies, improved technology integration, and organizational tactics.

2025, Deep Science Publishing

2025, Vol. 18 No. 4 (2025): July - August

The aging population in Thailand is proliferating, with significant effects on social, economic, and health systems. This study aims to investigate the impact of financial fraud on senior females' mental health in the Phra Khanong district of Bangkok, Thailand. It identifies the types of fraud experienced by seniors and their psychological consequences, primarily focusing on stress, anxiety, and depression. The researcher used mixed methods to explore senior females' experiences with financial fraud. Data collection tools included the Kessler Psychological Distress Scale (K-10) to assess the level of distress, General Anxiety Disorder-7 (GAD-7) to measure anxiety levels, Patient Health Questionnaire-9 (PHQ-9) to evaluate depression severity, and in-depth Interviews to provide qualitative insights into personal experiences with fraud. The findings indicated a high prevalence of financial fraud among the sample, with the most common phishing scams via text messages and social media (70%). Victims of financial fraud exhibited higher levels of distress, anxiety, and depression compared to non-victims. Specifically, the mean distress score was markedly higher in the fraud group, highlighting the severe emotional impacts of financial exploitation. These findings call for urgent legislative action, increased public awareness, innovation, and collaboration with financial institutions to protect this vulnerable population from financial exploitation. Future research should expand the geographical scope of the study and incorporate longitudinal designs to understand better the long-term psychological effects of financial fraud on seniors. By prioritizing the psychological well-being of elderly individuals, society can better protect one of its most vulnerable populations from the detrimental effects of financial exploitation.

2025

This study is driven by two objectives; First, the study seeks to predict the effect of forensic accounting in Ghana using the machine learning models. Second, it adopts the feature importance ranking algorithm to examine the impact of forensic accounting in Ghana. Accounting leaders have been urging educators worldwide to teach accounting students about fraud and other forensic accounting services, including how to conduct business valuations, provide litigation support, and provide expert witnessing services, as well as how to prevent, detect, and investigate fraud (Alshurafat et al., 2020). Economic and financial crimes (EFCs) are a global issue affecting all firms and economies; thus, there is a high demand for prevention and uncovering these crimes by institutions and nations. (Dote-Pardo & Severino-González, 2025; Ocansey, 2017). However, forensic accounting, coined in 1946, emerged to combat these crimes and improve investor confidence (Polycarp, 2019). In order to stop financial losses and preserve the integrity of the financial system, forensic accountants are essential (Emmanuel et al., 2018). To track fraudulent activity, find offenders, and comprehend fraud techniques, they analyze data. Their responsibilities now include creating plans to lessen online financial fraud. In digital information sectors, forensic accountants help manage internal issues, increasing operational effectiveness and efficiency and lowering the risk of fraud (Manning, 2010; Tuli & Thaduri, 2023). Moreover, financial fraud has become more difficult due to the emergence of new technologies like blockchain and cryptocurrencies. As forensic accountants use their specialized skills to investigate and settle fraud cases, they are becoming more and more involved in instances involving these technologies. In conclusion, forensic accounting plays a complex and ever-changing role in the fight against digital financial fraud. In the digital age, forensic accountants are leading the charge in identifying, discouraging, and looking into financial crimes (Patel, 2019; Saluja et al., 2024).

2025

2025, World Journal of Advanced Engineering Technology and Sciences

Artificial intelligence is transforming the financial services industry through revolutionary applications in risk management and fraud detection. This transformation extends beyond incremental improvements to fundamentally reimagine core financial processes, enabling institutions to process vast quantities of data, identify complex patterns, and make decisions with unprecedented speed and accuracy. AI-driven systems have evolved risk assessment beyond traditional statistical models by analyzing billions of variables simultaneously and detecting subtle correlations invisible to human analysts. In fraud detection, sophisticated anomaly detection algorithms establish individualized behavioral baselines for each customer, dramatically reducing false positives while preserving legitimate transactions. These systems identify fraudulent patterns in real-time, detect novel schemes, and recognize coordinated fraud rings with remarkable precision, translating directly to significant reduction in fraud losses and increased transaction volumes. Behavioral analytics has created unparalleled visibility into customer financial patterns, supporting both enhanced fraud prevention and hyper-personalized service offerings. As these technologies continue to mature, financial institutions must balance innovation with ethical considerations and regulatory compliance, recognizing that trustworthiness represents a powerful competitive advantage in an increasingly algorithm-mediated landscape.

2025, Estrategia Organizacional, 14 (1)

Introducción. El objetivo de esta investigación es evaluar la viabilidad técnica y organizacional para el montaje de una fábrica de jeans en Riohacha, Colombia. Se parte de la necesidad de aprovechar oportunidades locales de mercado en el... more

2025, IJCET_16_03_031

Digital innovations, including cloud-native technologies, are continuously transforming the financial services sector. Financial institutions are embracing cloud-native adoptions to deploy artificial intelligence and machine learning models that address their operational threats, deficiencies, and compliance. Nonetheless, these adoptions introduce significant cybersecurity concerns that require the integration of validated and adaptable frameworks like the NIST cybersecurity framework (NIST CSF). By employing the systematic-narrative hybrid literature review methodology, this paper examined the role of the NIST CSF in the cloud-native technology adoption in the financial services sector. Literature from academic databases such as Google Scholar, Scopus, ResearchGate, IEEE Xplore, and ScienceDirect, as well as NIST and IBM publications, was extracted for the review. The review concluded that the NIST CSF is a structured guideline for handling cybersecurity risks in the financial services sector, with the potential to improve scalability, compliance, resilience, and governance. By balancing regulatory compliance with dynamic security infrastructure development, the framework can also be adapted for unique challenges in decentralised environments. Organisational culture and management support were also identified as factors that enhance the effectiveness of NIST CSF integration for cloud-native adoptions.

2025, INTERNATIONAL JOURNAL OF EMERGING RESEARCH IN ENGINEERING AND TECHNOLOGY

Combining blockchain with AI is heavily transforming digital banking by facilitating intelligent, secure, and realtime decision-making processes. While financial institutions move away from legacy systems toward data-driven platforms, there is a growing need for real-time BI. Most transitional BI tools are thus limited by the presence of centralized data silos, slow data pipelines, and lack of transparency. In comparison, blockchain ensures a decentralized tamper-proof ledger infrastructure that gives assurances of data integrity, traceability, and auditability, whereas AI offers tools for extracting actionable insights such as predictive analytics, anomaly detection, and natural language processing. In pausing this study turns its focus on the synergistic integration of blockchain and AI toward real-time BI framework developments within digital banking ecosystems. A multi-layered architecture is thereby proposed wherein blockchain captures, validates, and stores transactional and behavioral data, whereas a layer of AI modules sit atop this secured data layer to generate intelligent patterns in real time. This research puts forward supervised learning models such as XGBoost and LSTM for fraud prediction and customer segmentation, while smart contracts trigger compliance workflows and rule-based alerts. Explainable AI techniques (e.g. SHAP, LIME) are also integrated for purposes of interpretability and regulatory compliance. Results indicate that fraud detection accuracy has been improved to 96%, latency to real-time insight generation has dropped substantially to a negligible level, and trust in AI results has been strengthened by the transparency of blockchain logging. Case studies of customer behavior analytics, transaction anomaly monitoring, and credit scoring show how this integrated approach outperforms traditional data infrastructures. Besides, this work has put forward other discussions on challenges in implementation such as interoperability, data privacy, computational costs, and regulatory acceptance. This research contributes to the fast-evolving discourse on digital transformation in finance, offering a scalable, secure, and interpretable blueprint for next-generation banking systems, which will take advantage of blockchain and AI in providing real-time intelligence.

2025, Leveraging ICT and Big Data Analytics for Transformative Business Solutions in Uganda

The Fourth Industrial Revolution has ushered in an era where Information and Communication Technology (ICT) and Big Data Analytics are pivotal in reshaping business landscapes globally. In Uganda, the rapid adoption of ICT presents both challenges and significant opportunities for businesses to enhance decision-making, improve operational efficiencies, and foster innovation. Despite these opportunities, challenges such as inadequate infrastructure, limited digital literacy, and high implementation costs remain significant barriers, particularly for small and mediumsized enterprises (SMEs).

2025, International Journal of Engineering, Management and Humanities

Fraud detection for online banking continues to be an important challenge given the sophistication of frauds and the high-dimensionality of online transactions. Classical machine learning and deep learning algorithms, such as rule-based classifiers, decision trees, and neural networks, find it hard to identify evasive fraudulent patterns in real time with high recall and precision. In this paper, we introduce a Deep Convolutional Autoencoder (DCAE)powered anomaly detection framework that utilizes reconstruction error to detect fraudulent transactions. The model extracts hierarchical transaction features automatically without feature engineering, increasing fraud detection efficiency. The PaySim dataset is used to evaluate the model with a performance measure of 98.7% accuracy, which is much better compared to conventional methods such as decision trees and neural networks. The results authenticate that the proposed DCAE model has a scalable, effective, and real-time fraud system with significant elimination of false positives and enhanced security for online banking transactions.

2025, International Journal of Engineering Research and Science & Technology

Personalized medicine is rapidly advancing with deep learning and predictive analytics, starting from using electronic health records to improve clinical decision-making. These technologies advance disease prognosis, treatment customization, and management of healthcare resources, taking the sector toward a proactive approach. Though it looks forward to optimizing patientcentric decision support via DL and predictive analytics in improving clinical decision-making, personalization of treatment, health outcomes forecasting, optimal resource allocation, and patient satisfaction, data integration issues and privacy, as well as issues of interpretability are the challenges on the way. This will integrate deep neural networks with predictive models for the analysis of structured and unstructured EHR data for better accuracy using feature engineering, data augmentation, and hyperparameter tuning. The performance evaluation is done on the basis of real-world patient data, thereby leading to significant improvements in prediction reliability, treatment personalization, and efficiency of decision-making over traditional models. Despite implementation challenges, these technologies promise improved treatments, reduced healthcare costs, and better patient outcomes. Future efforts should emphasize broader integration and ethical considerations.

2025, International Journal of Information Technology & Computer Engineering

This study focuses on ECG monitoring and investigates how cloud computing, fog computing, and IoT can be used to create scalable and efficient healthcare solutions. Patients' ECG signals are continuously collected by IoT devices and analyzed locally at fog nodes, which ensures minimal latency and lessens the strain on the cloud. By processing data near the source, fog computing allows for quicker reaction times and instantaneous analysis and decision-making. Cloud computing enhances fog by offering large-scale storage, processing capacity, and robust machine learning models for analyzing huge datasets-all of which are essential for long-term storage and precise forecasting. ECG signals are used to identify abnormal heart conditions like arrhythmias or ischemia using machine learning-driven approaches like feature extraction and anomaly identification. This improves the precision of diagnosis and makes prompt actions possible. When compared to conventional systems, the system's 94% accuracy in real-time ECG analysis greatly increases anomaly detection rates and scalability. In addition to improving the system's scalability and efficiency, the combination of cloud, fog, and IoT also makes it possible for the system to manage large data streams with little latency. Cloud and fog computing together create new opportunities for healthcare systems to become more precise, responsive, and efficient, setting the stage for the future of digital healthcare.

2025, International Journal of Information Technology and Computer Engineering

This article investigates the integration of artificial intelligence(AI) in the health care industry ,with a particular emphasis on using the Turkish National AI Strategy and the AI CognitiveEmpathy Scale toimprove market performance andpatient satisfaction. AI-driven valuecreation in healthcare strives to improve patient outcomes, optimize resource use, and boostoverall healthcare delivery system efficiency. The report highlights the importance of AItechnology in transforming healthcare into a more personalized, efficient, and patient-centricsystem. The report, which aligns with Turkey's National AI Strategy, illustrates how AI canhelp Turkey achieve its goal of being a global leader in AI-powered healthcare solutions.Additionally, we evaluate the AI Cognitive Empathy Scale for its potential to enhance patient happiness by enabling AI systems to more accurately perceive and respond to human emotions.The research uses a mixed-methods approach that combines qualitative and quantitative data to assess the impact of AI-driven strategies on healthcare performance,demonstrating considerable gainsin patientcare,resource efficiency, and market competitiveness

2025, International Journal of HRM and Organizational Behavior

Artificial intelligence (AI) driven by machine learning has revolutionized the identification of financial fraud in Internet of Things (IoT) environments. This technique quickly and accurately identifies suspicious patterns in the vast and diverse data streams from IoT devices through the application of advanced algorithms, potentially identifying fraudulent activity. Using methods like anomaly detection and clustering, along with supervised and unsupervised learning that are trained on historical transaction data, artificial intelligence systems are able to distinguish between legitimate and fraudulent transactions with great accuracy in real time. In order to create trustworthy fraud detection models in Internet of Things environments, this study looks at the methodology, datasets, and assessment metrics that are necessary for adaptive learning through frequent retraining and automatic reaction mechanisms.

2025, Journal of current science

A potent strategy for organizing and evaluating massive amounts of data is to integrate big data, hashgraph, and cloud computing within the framework of the Kinetic methodology. Scalable resources are made available via cloud computing, enabling rapid and safe processing of large datasets. Better decision-making results from the extraction of insightful information made possible by big data analytics. Hashgraph technology, renowned for its quick and safe consensus process, guarantees operational effectiveness and data integrity. In addition to addressing issues with interoperability, scalability, and regulatory compliance, this study looks at how various technologies might be combined to enhance productivity, decision-making, and data security.

2025, International Journal of Recent Advances in Multidisciplinary Research

Financial fraud activities are a serious threat to the security and integrity of online banking systems. Traditional fraud detection approaches, such as rule-based and simple machine learning models, are not effective in detecting changing patterns of fraud and suffer from high false positive rates and scalability. To overcome these drawbacks, this research introduces BankSafeNet, a Dual-Autoencoder and Transformer-Based Anomaly Detection System for detecting financial fraud. The suggested framework utilizes a dual-autoencoder architecture to learn transaction patterns and identify anomalies, while a transformer-based classification model learns sequential relationships in transaction data. The system provides a fraud probability score and marks suspicious transactions for investigation. Measured on the PaySim dataset, the developed model records 99.45% accuracy, 99.54% precision, 99.37% recall, and 99.45% F1-score, performing much better than conventional fraud detection methods. The model also has a false positive rate (FPR) of 0.469% and a false negative rate (FNR) of 0.634%, which prove it to be highly resilient in terms of reducing false positives while its fraud detection correctness remains high. The findings demonstrate the effectiveness of BankSafeNet in furnishing an scalable, real-time fraud detection platform that complements financial security of digital transactions.

2025, International Journal of Information Technology & Computer Engineering

Background information: Secure document clustering is now more important than ever because to the exponential explosion of data brought forth by IoT systems in business, smart cities, and healthcare. In order to improve security and efficiency, this study suggests a system that combines Multivariate Quadratic Cryptography (MQC) with Secure Document Clustering (SDC) and Affinity Propagation (AP). Techniques include clustering with AP and encryption with MQC. The technology seeks to guarantee effective clustering while protecting data. The findings demonstrate that the suggested approach greatly increases accuracy, scalability, and security.

2025, International Journal of Science and Engineering Applications

Cloud storage, processing, and analysis of health data are, therefore, among the very central considerations of this study toward improving health decisions via data management. It develops the cloud framework for health data analysis and classification to increase predictive accuracy. Data collection begins with cloud platforms and develops toward storage of increasingly flexible, scalable databases for health. The data comes into pre-processing or cleaning and normalization steps for analysis; a Linear Discriminant Analysis (LDA) feature is extracted from the data while maintaining the attention of classdiscriminative information. A genetic algorithm in feature selection conducts feature relevance identification to improve performance. Finally, classification has been done using TabNet for better classification. This research study emphasizes a healthcare predictive model that claims high accuracy (98.7%), precision (98.6%), recall (99.1%), and F1 score (98.6). Throughput also increases with request rate values, but stabilizes at 0.5 req/s after 1.0 req/s, indicating diminishing returns, and thus is scalable and reliable with more comprehensive insights about health data and optimized predictions about outcomes of a patient so further decisions may be based on this information in patient healthcare.

2025, Journal of Science & Technology

Background information: Data sharing has been transformed by the Internet of Things' explosive expansion, yet security and privacy threats have increased. Methods: For safe IoT data sharing, this study combines isogeny-based hybrid cryptography with anisotropic random walks (ARW) and decentralised cultural co-evolutionary optimisation (DCCO). Our goals are to develop a safe model for IoT data sharing, optimise it using DCCO, and improve the security of data transfers. Results: With 97% overall accuracy and 96% data secrecy, the suggested approach performed better than conventional techniques. Conclusion: By maximising robustness and performance, this innovative method improves IoT security.

2025, International Journal of Advances in Engineering and Management

Data privacy and security have emerged as crucial issues in the healthcare insurance environment, owing to increasing reliance on digital data. Protecting sensitive customer data would be a fundamental requirement for compliance and risk management against threats of unauthorized access. The study is to develop a secure cloud-based solution to encrypt health insurance policyholder data, implement authentication mechanisms, and safely store the data in cloud storage for future analysis.Initially,data collection whereby all data integral to insurance may be collected. The system gives strict access control via Zero Trust Authentication such that only authorized people can get access to the data. Next, high-level encryption of data using Elliptic Curve Cryptography takes place, after which the data is safely stored with the intent of scalable and efficient management in Cloud Storage. The outcomes proved that the cryptographic overhead is 30 units while achieving a very low cryptographic failure (5%). Moreover, the cloud computing latency has been optimized, wherein the response time enhances from 375 ms to 100 ms by the 10th time step, The contribution of this study lies in advancing an integrated approach for healthcare insurance data management that is complete, secure, and based on modern encryption and authentication techniques and provides an understanding of how cloud storage solutions could effectively manage sensitive data with high security.

2025, International Journal of Applied Science Engineering and Management

Background The emergence of the Internet of Things (IoT) generates large quantities of data and calls for scalable storage and real-time analytics technologies for the complexity of IoT in the smart city, healthcare, etc., industries. Methods This paper presents an approach for the analysis of IoT real-time data, using IBM Cloud Object Storage, Azure Stream Analytics, mesh networks, and Hidden Markov Models (HMMs). Objectives The main objective of this study is to enhance the scalability, efficiency, and predictive analytics of IoT systems by integrating cloud-based storage, real-time analytics with high throughput, and decentralized mesh networks for improved performance. Results The proposed method was 92% accurate, 90% efficient, and 93% scalable, outperforming current methods through increased real-time framing, reliable networking, and improved predictive power via HMMs. Conclusion This hybrid method serves as a scalable and deployable solution to any IoT system that is advantageous for real-time applications in smart cities, healthcare, and industrial automation.

2025, international journal of humanities and social science and management

This paper describes a Secure Health Care Heart Disease Monitoring and
Classification System which utilizes advanced deep learning techniques combined with the Internet of Medical Things (IoMT) devices and cloud storage for a successful prediction of heart diseases. While convolutional neural networks (CNNs) are used to classify the signal, the feature extraction from the signal is done via wavelet transform. It efficiently
detects and monitors many disorders of the heart, such as arrhythmias, myocardial infarction, and heart valve diseases, in real-time. The suggested framework promises a powerful combination of secure management of sensitive medical information through MAC verification, along with cloud hosting. This combination allows legitimate personnel access while denying any ulterior access. This approach addresses modern challenges in health systems on data integrity, privacy, and
resource management while also expediting and bolstering the diagnostic confidence. Its accuracy against conventional methodologies is, however,
reported at a striking 98.32%.

2025, International Journal of HRM and Organizational Behavior

Financial fraud is an important challenge of digital banking that requires strong and dynamic detection models. In this paper, a Variational Autoencoder (VAE)-Gated Recurrent Unit (GRU) with Attention Mechanism for fraud detection in financial transactions has been proposed. The model utilizes VAE for detecting anomalies in an unsupervised manner, GRU for extracting sequential patterns of dependencies in transactional patterns, and Attention Mechanism for improving feature selection. The suggested method is tested on a large-scale Fraudulent Transactions Dataset of 6,362,620 transactions, showing 99.25% accuracy, 99.62% precision, and 98.88% recall. Additional testing with AUC-ROC (0.9933) and Average Precision Score (0.9933) validates its efficiency in identifying fraudulent activities. Comparative results indicate the model's superiority over conventional machine learning methods in fraud detection. The research highlights the necessity of combining deep learning methods with attention mechanisms to increase the accuracy of fraud detection and reduce false positives.

2025, The American Journal of Engineering and Technology

In this study, we present a deep learningbased approach for real-time credit card fraud detection in banking systems, with a primary focus on Long Short-Term Memory (LSTM) networks. Using a highly imbalanced credit card transaction dataset, we implemented comprehensive preprocessing, feature engineering, and model evaluation strategies to enhance the detection accuracy. Our experimental results reveal that the LSTM model significantly outperformed traditional machine learning algorithms such as Logistic Regression, Decision Tree, and Random Forest. The LSTM achieved an accuracy of 99.38%, precision of 99.40%, recall of 99.22%, and F1-score of 99.31%, demonstrating its superior capability to detect fraud while minimizing false positives. Through comparative analysis, we establish that deep learning not only improves predictive performance but also adapts better to temporal patterns inherent in financial transactions. This research underscores the transformative potential of AI-driven fraud detection in modern banking infrastructures, ensuring enhanced security, operational efficiency, and customer trust.

2025, Journal of Ubiquitous Computing and Communication Technologies

Traditional healthcare systems have difficulties such as delayed diagnosis, resource
constraints, and data security issues, particularly during pandemics. Lightweight CNNs,
capsule networks, and DAG-based blockchain alternatives are all included in a next
generation healthcare system to improve diagnostic precision, scalability, and decentralized
data security. With GANs creating synthetic datasets for training, this method uses DAGs for
safe and scalable data sharing, lightweight CNNs for feature extraction, and capsule networks
for spatial representation. The real-time performance and interoperability of a modular design
are confirmed by measurements for accuracy, sensitivity, and latency. In terms of safe data
sharing and real-time pandemic detection, the suggested system outperformed traditional
techniques with 99.9% data integrity, 96.4% accuracy, 97.1% sensitivity, 23.3 ms latency,
and 1200 TPS scalability.

2025, International Journal of Engineering and Science Research

Background information: Secure, low-latency data sharing has become more difficult as a result of the Internet of Things' explosive expansion. This paper suggests a fog computing system that uses Federated Byzantine Agreement (FBA) for safe and scalable data sharing, Directed Acyclic Graph (DAG) protocols, and Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and Firefly Algorithm for optimization. By addressing latency and security, the method guarantees effective data sharing. In a variety of IoT scenarios, the results demonstrate increased throughput, security, and decreased latency. Methods: The project incorporates DAG protocols for data routing, FBA for consensus, and CMA-ES and Firefly Algorithm for optimization in fog computing settings. Using a decentralized, fault-tolerant architecture to optimize resource utilization and improve security, these methods guarantee reliable, low-latency data transfer. Objectives: The goal of this research is to create a framework for safe and effective IoT data sharing based on fog computing. Using CMA-ES and the Firefly Algorithm for optimization, it aims to lower latency and improve scalability. It also employs FBA for strong consensus processes and DAG protocols for organized data routing, guaranteeing data integrity and defence against malevolent attacks. Results: Data sharing was significantly enhanced by the suggested paradigm, which also showed increases in throughput of 94%, energy efficiency of 85%, and latency reduction of

2025, international journal of humanities and social science and management

The increasing availability of healthcare data through electronic health records (EHRs) has opened up opportunities for advanced abnormality detection in patient health monitoring. However, existing systems face significant challenges, including the processing of large volumes of realtime data, incomplete datasets, and inaccurate predictions for healthcare professionals. This paper proposes a robust healthcare abnormality detection model that aims to accurately identify abnormal patterns in patient health data obtained from cloud. The proposed process begins with collecting data from cloud storage, followed by preprocessing steps such as imputing missing values and removing outliers using Z-score standardization to ensure data integrity. Feature extraction is then performed using Fast Fourier Transform (FFT), which converts timedomain signals into the frequency domain, capturing periodic patterns that may indicate abnormalities. The extracted features are passed to a Gated Recurrent Unit (GRU) and Deep Neural Network (DNN) hybrid model for abnormality detection. The model achieves outstanding results, with an accuracy of 99.45%, precision of 99.50%, recall of 93.00%, and F1-score of 97.00%. Furthermore, the system shows a reduction in error rate from 0.275 to 0.125, indicating performance improvement with parameter optimization. These findings demonstrate the system's scalability, efficiency, and potential for enhancing healthcare abnormality detection in cloud-based environments.

2025, International Journal of Information Technology & Computer Engineering

Financial fraud detection continues to be an important challenge as a result of changing fraud schemes and highdimensional transactional information. This work introduces TransSecure, a Transformer-based anomaly model with self-supervised learning incorporated for financial fraud detection. The model employs a Masked Transaction Model (MTM) for pretraining with masked financial data to enhance its capacity to detect fraudulent activities. Self-attention mechanisms allow for the identification of short-term and long-term fraud patterns by modeling intricate dependencies in transaction sequences. The approach is tested on a large-scale Fraudulent Transactions Dataset with 99.31% accuracy, 99.54% precision, 98.08% recall, and an AUC-ROC of 0.9934. Experimental results show that TransSecure effectively minimizes false positives and negatives compared to conventional machine learning and deep learning models. This research demonstrates the power of self-supervised Transformers in detecting financial fraud and offers insights into actual fraud prevention methods.

2025, International Journal of Information Technology and Computer Engineering

Background: Cloud computing (CC) and artificial intelligence (AI) are causing a rapid evolution in healthcare, meeting the requirement for accurate and effective disease diagnosis and management through wearable IoT devices and sophisticated algorithms. Objective: To develop a BBO-FLC and ABC-ANFIS system that works together for better disease prediction accuracy and real-time monitoring. Methods: Implemented on a scalable cloud architecture, the system combines IoT-enabled sensors for data gathering, ABC for feature optimization, BBO for fuzzy rule refining, and ANFIS for disease categorization. Results: The suggested solution outperformed conventional techniques with 96% accuracy, 98% sensitivity, and 95% specificity at a 60-second computation time reduction. Conclusion: The precision, scalability, and real-time healthcare applications for complicated disease prediction and monitoring could be greatly improved by this integrated system.

2025, International Journal of Information Technology & Computer Engineering

Background information: Monte Carlo simulations, Deep Belief Networks (DBNs), and Bulk Synchronous Parallel (BSP) processing are used in the suggested secure cloud-based financial analysis system to increase the effectiveness of risk prediction and financial modeling. The system uses cloud infrastructure for precise financial forecasts to guarantee scalability, security, and high-performance data processing. Computational time is greatly decreased by parallel processing, and data security is preserved by encryption, facilitating sound decisionmaking in intricate financial contexts. Methods: The system combines Monte Carlo simulations for forecasting risks, DBNs for identifying patterns, and BSP processing to enhance computational efficiency within a cloud setting. Data that is encrypted is processed over multiple cloud nodes, improving security and scalability. This integration enables simultaneous processing of multiple simulations, thus enhancing the speed and precision of financial analysis. Objectives: By combining Monte Carlo simulations, DBNs, and BSP processing, this work seeks to improve the efficiency of financial models and provide a safe, cloud-based financial analysis system. The solution is designed to safely manage huge datasets in a cloud environment that is scalable. The system aims to shorten calculation times by utilizing parallel processing, guaranteeing precise financial forecasts, risk assessment, and trustworthy decisionmaking. Results: The suggested system performs financial analysis jobs more accurately and efficiently. Strong security and scalability are offered by encrypted data management and parallelized simulations. Performance measures show notable gains in recall, accuracy, and precision over conventional techniques, with BSP processing improving scalability. For crucial financial decision-making, this method facilitates quick and safe data analysis. Conclusion: The suggested system performs financial analysis jobs more accurately and efficiently. Strong security and scalability are offered by encrypted data management and parallelized simulations. Performance measures show notable gains in recall, accuracy, and precision over conventional techniques, with BSP processing improving scalability. For crucial financial decision-making, this method facilitates quick and safe data analysis.

2025, International Journal of Information Technology & Computer Engineering

This paper proposes an advanced adaptive access control system for Smart Healthcare and Cloud Systems (SHACS) that combines Markov Models, Topological Data Analysis (TDA), and feature optimization. The system integrates Markov Chains' probabilistic modeling with the structural insights provided by TDA to dynamically assess user access requests. Markov Models predict future access patterns by examining historical data, while TDA analyzes the structural patterns of user interactions to identify anomalies and vulnerabilities in the cloud system. This fusion of methods enables the system to not only detect deviations from expected behavior but also predict and mitigate potential threats in real-time. In order to keep the system sensitive to changing security threats, the adaptive mechanism makes use of feedback loops to continuously update and

2025, International Journal of Research in Engineering Technology

Background Cardiology has struggled to operationalize patient-specific care pathways. Decision trees, an AI method that offers interpretable decision-making are used in crowdsourcing by harnessing collectively intelligent inputs. They partner to identify innovative ways for advancing personal medication solutions aimed at more individual and efficient cardiac outcomes. Methods This study employed crowdsourcing to elicit patient-centric data and decision trees (DTs) as an analytical approach for optimizing therapeutic pathways in cardiology. We evaluated a set of performance indicators to compare the standard against new methods, such as data accuracy and prediction accuracy. Objectives A study will be designed to evaluate the application of crowdsourcing methodologies in different patient data-mining, care optimization with decision trees, and improving prediction accuracy along with cost reductions and optimizing patient happiness using cardiology. It will also highlight the success of this comprehensive approach to healthcare delivery. Results Our approach outperforms the existing methods in terms of 93% data & prediction accuracy, 85% cost saving, and finally achieved patient satisfaction level as around more than one is equal to ninety. play the process of crowdsourcing and optimizing decision trees on clinical pathways significantly increased total resource efficiency with improved care personalization in cardiology. Conclusion Enabling crowdsourcing through decision trees better cardiology treatment pathways by increasing accuracy, reducing costs, and enhancing patient experience. Together, these dual-mode system strategies promise scalable and patient-centered healthcare solutions with the possibility of wider applications in hospitals.

2025, international journal of humanities and social science and management

As banking system financial fraud becomes more sophisticated, it necessitates sophisticated fraud detection and prevention methods in real time. DeepBankGuard, a hybrid deep learning framework proposed in this paper, is specifically designed to detect bank fraud using Variational Autoencoder (VAE) and Attention-based Bidirectional Long Short-Term Memory (BiLSTM) networks. The VAE is used to extract compressed feature representations from financial transactions, while the BiLSTM model is used for detecting long-range sequential patterns in user transaction behaviour. An attention mechanism is used to emphasize the most significant features that result in fraudulent behaviour. The model achieves outstanding performance with 99.56% accuracy, 99.61% precision, 99.51% recall, and 99.56% F1-score in the test set, and its effectiveness in identifying fraud with negligible false positive and false negative rates (0.391% and 0.489%, respectively). The technique beats traditional techniques, and it results in a scalable, adaptive, and real-time fraud detection system for secure banking applications. The VAE and BiLSTM model pairing constitutes a robust model for solving fraud detection problems in the dynamic financial world.

2025, International journal of modern electronics and communication engineering

Background Information: This study amalgamates blockchain technology with artificial intelligence and Sparse Matrix Decomposition methodologies to tackle data management issues within Human Resource Management. Conventional HRM systems encounter constraints in security, scalability, and decision-making efficacy, particularly when dealing with extensive, partial datasets. Blockchain guarantees data security, whereas AI offers predictive analytics to enhance HRM functions. Objectives: The project seeks to create a secure, scalable, and effective HRM data management system by integrating blockchain's immutability with AI's predictive skills and Sparse Matrix Decomposition to manage extensive, sparse information and improve decision-making in HR processes. Methods: A prototype system was created utilizing blockchain for decentralized data storage, AI-driven predictive control for human resources trends, and Sparse Matrix Decomposition to handle extensive, incomplete datasets. The system's performance was evaluated based on critical parameters including security, scalability, processing time, and storage efficiency. Results: The proposed solution markedly enhanced data security (0.99), scalability (0.95), storage efficiency (0.96 GB), and prediction accuracy (0.95), surpassing alternative methods in HR data management. Conclusion: Integrating blockchain, artificial intelligence, and Sparse Matrix Decomposition yields a resilient and scalable system for Human Resource Management. It improves data security, prediction accuracy, and system efficiency, providing a revolutionary method for managing extensive HR datasets and enhancing decision-making processes.

2025, International Journal of Computer Science Engineering Techniques

Background Information: In the contemporary digital environment, the secure management of employee data is an escalating concern owing to rising cyber risks and data breaches. Conventional centralized systems frequently encounter challenges related to security and scalability, requiring sophisticated methodologies that incorporate blockchain, artificial intelligence (AI), and machine learning (ML) to effectively safeguard critical employee data.

2025, International journal of modern electronics and communication engineering

By using a hybrid approach combining robotic automation, Autoencoder-LSTM models and fuzzy cognitive maps (FCMs), this paper introduces an IoMT-based technology that offers the best Intelligent System for CKD prognosis. IoMT devices are utilized to capture real-time health data for continuous patient monitoring such as blood pressure and serum creatinine. Autoencoders are used to down-sample the data, while thenceforth sequence prediction is carried out by LSTM networks. FCMs are employed to stage the phases of CKD using a complex medical scenario and decision-making simulations. Robotics automation is made easy when processed real-time for better-quality management and accuracy. The method respects early CKD detection and successfully outperforms the conventional model by 98.96% accuracy. Objectives: The primary goal is to develop the CKD stage prediction accuracy utilizing IoMT and robotic automation while with Autoencoder-LSTM models are employed for identifying disease stages which in turn, assist to simulate complex medical decisions associated with the aforementioned prediction using FCMs. Methods: IoMT data being collected real-time and analysis to be done using autoencoders for dimensionality reduction and feature selection. FCMs predict disease state while LSTM models predict the trajectory of CKD. Robotic automation allow real-time data handling efficient.