1393 research outputs found
Sort by
Business intelligence adoption in higher education: the role of data-driven decision-making culture and UTAUT
No full textPurpose
While “business intelligence” (BI) has demonstrably improved decision-making across various industries, its application in “Higher Education Institutions” (HEIs) remains under-explored. HEIs collect vast amounts of data on students, faculty and operations. The purpose of this paper is to investigate the significant variables affecting the adoption and utilization of BI in HEIs, aiming to bridge this gap in knowledge. Design/methodology/approach
Drawing upon the theoretical foundation provided by the “Unified Theory of Acceptance and Use of Technology” (UTAUT) framework, this research developed a novel conceptual model integrating a context-specific variable: “data-driven decision-making culture” (DDMC). To investigate the effect of these factors within Jordanian HEIs, a cross-sectional Web-based questionnaire was administered to 427 individuals holding various management positions. Findings
“Structural equation modelling” (SEM) analysis demonstrated that perceived “performance expectancy” (PE), “social influence” (SI) and “facilitating conditions” (FC) significantly affect behavioural intentions to use BI systems. Furthermore, these results suggest that an organizational culture emphasizing data-driven decision-making acts as a significant antecedent within the UTAUT model for BI technology adoption. Originality/value
This investigation holds significant implications for the practical application of BI technology within HEIs. These insights are useful for the HEIs stakeholders in the development, design and provision of services, as well as policymakers in the same sector. These key findings can help inform the development of effective strategies and BI systems for HEIs. These advancements have the potential to broaden user engagement in BI systems within the HEI community
Using Social Media Data to Predict Mental Health Issues: A Tertiary Study
No full textAddressing the pervasive issue of mental illnesses in the U.S. necessitates innovative approaches. This study explores the potential of social media platforms as valuable sources for detecting mental health issues, leveraging the spontaneous and open expression of users\u27 thoughts and feelings. Previous research has applied machine learning techniques to social media data to predict mental health states, which this study aims to expand by providing a holistic view of the strategies used for identifying mental health concerns through social media analysis. Our research questions focus on the strategies for utilizing social media data, the efficacy of these strategies, the challenges faced, and the broader implications for healthcare delivery. Employing a tertiary investigation approach, we review secondary studies to identify trends and synthesize findings, aiming to offer comprehensive insights and guide future research in mental health service delivery through social media engagement
The Impact of the Internet of Things in Healthcare Delivery: A Systematic Literature Review
No full textThe increasing demand for efficient healthcare has spurred the adoption of IoT technologies, promising reduced costs and improved outcomes. This study addresses the integration of IoT in healthcare, focusing on patient-centered applications across prevention, diagnosis, and treatment. We explore three research questions concerning the main IoT applications, their drivers and challenges, and their impact on healthcare delivery. This systematic literature review synthesizes evidence from the literature to identify trends and mappings in IoT applications. The review includes a comprehensive framework for IoT in healthcare, enhancing patient engagement and care delivery. The study is structured first to define IoT paradigms and technologies, followed by proposing a healthcare framework, describing our methodology, and discussing the implications of our findings on future healthcare innovations
Spatiotemporal Deep Learning for Land Cover and Impervious Surface Mapping with Satellite Imagery
Get PDFThis research explores the dynamics of land cover classification using remote sensing time series data, emphasizing the need for efficient monitoring and resource management on Earth’s surface. With advancements in computational power and analytical methods, deep learning techniques, including Convolutional Neural Networks (CNNs) and Transformer neural networks, have emerged as state-of-the-art approaches for automating and operationalizing land cover classification at regional and global scales. This study introduces two distinct methodologies for land cover time series classification: Spatial Recognition and Temporal Alignment (SpaRTA) and Land Cover Artificial Mapping System (LCAMS). SpaRTA employs a U-Net architecture coupled with a Transformer encoder for effectively generating annual classifications and ensuring temporal alignment, outperforming comparable methods in terms of both validation and independent test datasets. LCAMS builds upon SpaRTA by integrating regional model fine-tuning, ensemble modeling, change detection, and multitask learning to enhance its scalability and generalization capabilities. Key findings indicate that both methodologies achieve high levels of temporal and spatial consistency, comparable to legacy products like the National Land Cover Database (NLCD) and Land Change Monitoring, Assessment, and Projection (LCMAP), while expediting product generation and reducing latency. Despite their strengths, challenges remain, including the inherent difficulties of inter-annual consistency and the reliance on specific data sources, which may limit performance. Future work should focus on improving model architectures, incorporating intra-annual information, and enhancing forecasting methods. Ultimately, this research demonstrates the significant potential of deep learning in automating land cover analysis, paving the way for scalable solutions in environmental monitoring and resource sustainability
Malware Classification Using Weighted Control Flow Graphs
Get PDFMalware remains a primary cybersecurity threat because traditional signature-based detection methods have difficulty matching the pace of evolving malicious code implementing complex obfuscation methods to evade detection. The current detection methods fail to identify new malware variations, which expose systems to damaging cyber attacks that result in major security breaches and delayed incident response capabilities. This research investigates the speed-performance gap between signature-based detection and control flow graph behavioral understanding through developing Weighted Control Flow Graph (WCFG), which merges structural program analysis with signature-based detection capabilities. The design science research method enabled this study to create two main research outcomes: a complete WCFG dataset and an enhanced machine learning-based malware detection system. The research methodology encompasses a PE feature extraction from the PEMML dataset, followed by CAPA signature matching for identifying malicious functions, then generating control flow graphs through Radare2 before combining data with signature-based weights. The processed dataset consisted of samples divided between five malware families, including Zbot, Locker, Mediyes, Winwebsec, and ZeroAccess, as well as benign software, which underwent XGBoost classification with SMOTE Tomek balancing and Random Forest feature selection. The WCFG method reached 90% accuracy in classification, which outperformed the unweighted control flow graph-only method by 4% because it achieved 86% accuracy. The weighted model demonstrated better performance with precision and recall for malicious and benign samples. The SHAP analysis established that signature-based weighting features played a major role in determining classification outcomes, proving the effectiveness of the integration approach. The research findings deliver major practical value to cybersecurity defenders through improved automated malware triage systems that reduce analyst time waste from false positives while minimizing threats that evade detection due to false negatives. The WCFG methodology presents a deployable solution that unites static analysis speed with improved detection precision to fulfill the essential requirement for flexible malware classification systems operating in advanced threat environments
Performance Measurement of BGP Extended with De-Facto Network Identification
Get PDFThe Internet relies on the Border Gateway Protocol (BGP) to share reachability information across the globe. This protocol has previously, intentionally or unintentionally, been misconfigured by its users. Although there are solutions proposed to improve the security of BGP, the adoption of those solutions is not universal. This study measured the performance impact of implementing a network identification mechanism in BGP messages by appending signature information within IPv6 extension headers. There was a statistically significant impact on the performance of the BGP speakers under test and further research is needed to determine if such an impact is worth it for the potential security benefits of easily identifying networks
Generative AI for Synthetic Data Creation: Building Mastery-Focused Educational Datasets
Get PDFThere is no easily-available dataset for mastery-focused education, where mastery replaces grades while accurately reflecting student performance. Student data is restricted due to privacy & security concerns. One K-12 app was recently discovered selling unmasked data on millions of students Synthetic datasets may solve this by providing utility, privacy preservation, scalability, customization, variability, and resistance to reverse-engineering Techniques used included autoencoders, variational auto-encoders (VAE), generative adversarial networks (GAN) , and copulas combined with GANs.https://scholar.dsu.edu/erposters/1012/thumbnail.jp
Scalable and Ethical Insider Threat Detection through Data Synthesis and Analysis by LLMs
Get PDFInsider threats wield an outsized influence on organizations, disproportionate to their small numbers. This is due to the internal access insiders have to systems, information, and infrastructure. Signals for such risks may be found in anonymous submissions to public web-based job search site reviews. This research studies the potential for large language models (LLMs) to analyze and detect insider threat sentiment within job site reviews. Addressing ethical data collection concerns, this research utilizes synthetic data generation using LLMs alongside existing job review datasets. A comparative analysis of sentiment scores generated by LLMs is benchmarked against expert human scoring. Findings reveal that LLMs demonstrate alignment with human evaluations in most cases, thus effectively identifying nuanced indicators of threat sentiment. The performance is lower on human-generated data than synthetic data, suggesting areas for improvement in evaluating real-world data. Text diversity analysis found differences between human-generated and LLM-generated datasets, with synthetic data exhibiting somewhat lower diversity. Overall, the results demonstrate the applicability of LLMs to insider threat detection, and a scalable solution for insider sentiment testing by overcoming ethical and logistical barriers tied to data acquisition
PyRHOH: A meta-learning analysis framework for determining the impact of compilation on malicious JavaScript identification
No full textAutomated identification of malicious JavaScript is a core problem within modern malware analysis. Code obfuscation is a common tactic used to evade detection. This obfuscation hinders both manual and automated detection methods, including neural network techniques. In order for these methods to effectively classify malware, it is beneficial to reduce the effects of obfuscation as well as to optimize the configuration and structure of the neural network to be well suited for the task. To overcome these challenges, we present a new framework: “PyRHOH” (“Python Repeatable Hyperparameter Optimization Harness”), a meta-learning framework that implements Bayesian optimization. The automated exploration and maximization of candidate hyperparameters using a Bayesian method adds structure and rigor to the selection of neural network hyperparameters, providing the assurance that an implemented design is optimal. In this study, we used the PyRHOH framework to determine optimal recurrent neural network architectures for the differentiation of malicious and benign JavaScript samples. We then used these neural networks to measure the degree to which compilation of raw JavaScript samples into bytecode via Google’s V8 JavaScript compiler affected classification accuracy. Classifying in-the-wild samples, compilation increased the detection rate from 76.88% to 95.84%. Among uniformly obfuscated samples, compilation increased the detection rate from an average of 76.76% to an average of 91.24% e compilation was performed. This shows that pre-processing JavaScript into compiled bytecode has a clear positive impact on neural network categorization
Interpreting Office Document Macros with Bi-Directional Transformer Models
No full textMicrosoft Office Document malware is prevalent today, even though some of the macros were developed 30 years ago. This paper provides a novel method to classify malicious office document macros with inter-pretability. Our approach combines the function semantics and keyword contexts to leverage the self-attention functionality of transformers. This research focuses on Bidirectional Encoder Representations from Transformers (BERT) model variants to evaluate and compare the accuracy and interpretability of transformer models in detecting office document macros. The model is evaluated on a dataset collected using Common Crawl. The results show that our method using BERT model variants provides more than 99% accuracy in detecting office document macros. Our research also shows that the BERT models can accurately attribute the classification outcome to the input tokens. Finally, we propose a novel solution to scan email attachments for malicious office document macros and provide attribution reports which not only labels the email as malicious but also attributes as to which tokens in the document are contributing positively towards the classification. This solution is integrated with Gmail as a workspace add-on. We hope that such solutions improve the trust of cyber security personnel in the model and threat detection mechanisms and fine-tune the model to eliminate false positives and biases