The generalizability of Machine Learning (ML) classifiers in threat detection and classification (TD-TC) is crucial, particularly in Software-Defined Network (SDN)-based Internet of Things (IoT) architectures, where dynamic network management and centralized control introduce unique security challenges. However, there is an ongoing debate regarding the role of Network Identifier Attributes (NIAs) in training and evaluating ML models using publicly available datasets comprising network behaviors. Therefore, this study investigates the influence of NIAs on model generalization by examining learning curves with cross-validation. The results show that Timestamp and Flow ID lead to memorization in classifier models when using the IoTID20 dataset. However, removing all NIAs still resulted in a relatively high classifier accuracy, with Decision Tree (DT) achieving 96.63
The expansion of mobile cloud computing (MCC) has led to a rapid evolution in how data and services are delivered across geographically dispersed infrastructures. While MCC enables flexibility, scalability, and ubiquitous access, it presents critical security and performance challenges. This paper proposes a comprehensive architecture that leverages Software-Defined Networking (SDN), Network Function Virtualization (NFV), and Cloud Access Security Broker (CASB) to enhance the resilience, scalability, and manageability of MCC systems. We introduce a layered architecture where SDN controls the network plane, NFV deploys dynamic security functions, and CASB enforces data governance in cloud services. Through extensive experimentation using Mininet and traffic simulation tools, we evaluate the system’s latency, throughput, scalability, and security performance under various traffic conditions. Results demonstrate that the proposed solution achieves latency overhead under 5ms, successfully mitigates more than 90
Complex interface structures hinder the manual classification of UI components. This paper introduces a prototypical approach for UI element classification that combines a custom browser extension for collecting structured metadata (e.g., DOM structure, CSS properties) with machine learning methods to assign UI components to subject-specific tasks. The approach enables reliable clustering without extensive manual annotation and supports the evolution of adaptive user interfaces. Evaluations in the domains of e-commerce and smart city applications demonstrate that a combination of structural and textual features provides sufficient discriminatory power. The method integrates with intelligent interaction systems and context-aware interfaces, reducing redundancy and addressing growing interface complexity. Similar to semantic segmentation techniques, it enables dynamic adaptation of interaction flows, contributing to improved user experience and functional interface design.
This systematic literature review investigates the influence of technological knowledge on user acceptance, intention, and utilization of Artificial Intelligence (AI), synthesizing findings from 25 empirical studies published between 2021 and early 2025. Employing an integrated Antecedents-Decisions-Outcomes (ADO) and Theories-Contexts-Methods (TCM) framework, the analysis reveals that research in this period predominantly utilizes quantitative SEM methods based on TAM and UTAUT frameworks, largely within higher education contexts. Key findings indicate that technological knowledge facets, particularly Objective Knowledge (e.g., AI literacy) and Self-Efficacy, are critical antecedents to AI adoption, though their influence is often indirect. These factors frequently shape adoption outcomes by impacting core perceptual mediators, notably perceived ease of use, perceived usefulness, and trust. The synthesis highlights significant gaps, including limited methodological diversity, a narrow contextual focus, under-explored theoretical perspectives beyond TAM/UTAUT, and insufficient investigation into actual AI usage behavior. Addressing these gaps through diversified research approaches is crucial for advancing a comprehensive understanding of AI adoption dynamics.
The integration of the Internet of Things (IoT) and Cyber-Physical Systems (CPS) has revolutionized industries by enhancing connectivity, automation, and efficiency. However, these advancements bring significant privacy challenges due to extensive data collection and interconnected ecosystems. This paper provides a comprehensive analysis of privacy challenges in IoT and CPS, categorized into technical, data management, user-centric, and ecosystem-wide dimensions. Key privacy threats, such as data breaches, inference attacks, device tracking, and insider risks, are explored through case studies and empirical evidence. Existing mitigation strategies, including encryption, privacy-enhancing technologies, access control mechanisms, and regulatory frameworks, are evaluated. Despite these efforts, challenges remain, such as scalability of privacy solutions, privacy concerns in emerging technologies like 5G and AI-driven IoT, and the balance between privacy, security, and usability. The paper emphasizes the need for interdisciplinary collaboration to address these challenges, fostering trust and ensuring the sustainable development of IoT and CPS technologies. Future research directions are proposed to develop scalable, user-centric, and adaptive privacy-preserving frameworks that align with evolving technological and regulatory landscapes.
