Early non-invasive screening technologies are a paramount priority in modern healthcare for identifying complex neurodevelopmental traits characterized by social, communicative, and behavioural challenges. Recent breakthroughs in computer vision and deep learning have established automated facial image analysis as a highly viable paradigm for objective clinical screening. This study introduces a robust, optimized weighted ensemble framework that integrates the complementary architectural strengths of convolutional and transformer-based networks for binary classification of these specialized facial trait profiles. The pipeline concurrently leverages EfficientNet-B5 for localized feature scaling, Data-Efficient Image Transformers (DeiT) for long-range global self-attention, and ConvNeXt for modernized, high-performance convolutional representations. To ensure generalization and counteract dataset selection bias, a strict stratified 5-fold cross-validation scheme is enforced, followed by an optimized out-of-fold weighted probability fusion mechanism. Experimental evaluation on a benchmark dataset demonstrates that the unified ensemble achieves a state-of-the-art classification accuracy of 95.67% and an ROC-AUC of 0.9788, significantly outperforming individual standalone baselines. These empirical results validate that bridging high-frequency local textures with low-frequency global contextual relationships minimizes predictive variance, offering an accurate, stable, and scalable computational screening solution for automated clinical environments.
Critical-infrastructure facilities increasingly rely on dense Wi-Fi deployments, which creates exposure to rogue access points, evil-twin impersonation, deauthentication attacks and hidden clients. The paper proposes Correlated Radio-SOC Localisation (CRSL), a framework that fuses distributed RF-sensor observations, wireless-activity features, indicators of compromise and SIEM/SOAR events. The method combines localisation based on multi-sensor measurements with temporal, identity and behavioural correlation of SOC events. The framework produces classified threat findings with an estimated uncertainty region and confidence score. The proposed architecture and evaluation protocol show how radio-event fusion can improve incident attribution and accelerate operational response in protected wireless environments.
In recent years, due to the widespread adoption of Agile and DevSecOps approaches to software development, the increasing number of abuses and malicious actions targeting information technology-based resources and the accelerating need for software applications to adhere compliance requirements with certain standards, policies and regulations (e.g., the European Union Cyber Resilience Act or industry standards HIPAA and GDPR), the guarantee of all three – quality, the level of security and com-pliance for the needs of developing and delivering software solutions has become of significant importance. The paper presents a holistic approach to automated monitoring of software quality, security, and compliance, which has been researched and developed for the needs of one of the most growing industries, namely the FinTech industry. It introduces SDLC Auditor, a software tool designed to automatically monitor, control, and evaluate processes throughout the entire Software Development Life Cycle (SDLC) within Agile environments with regard to adherence to software quality, security, and compliance requirements. The study concludes by highlighting the advantages of the SDLC Auditor in developing stable, secure, and reliable software systems.
The integration of generative artificial intelligence into engineering education raises the question of how to define permissible forms of its use in different learning activities without replacing the student’s personal contribution. The aim of this study is to propose a methodology for classifying these activities according to the permissible degree of generative AI use. The methodology considers the learning objective, expected outcome, required independence, risk of substitution, and possibilities for verification and traceability. As a result, a five-level framework, L0-L4, is developed, linking each level to permissible forms of use and control mechanisms. Its practical application is illustrated through a matrix of main learning activities and examples from engineering disciplines. The framework shows that generative AI can support learning when its use is transparent and controlled but should be restricted when it replaces the knowledge and skills being assessed.
The selection of communication technologies and protocols in machine-to-machine (M2M) systems is a complex engineering task influenced by latency, reliability, throughput and deployment constraints. This paper presents an experimental evaluation of HTTP/1.1, HTTP/2, RabbitMQ and Apache Kafka using an integrated AHP–TOPSIS decision model across LAN, WLAN and hybrid KNX–IP environments. Performance is assessed through latency, message loss and throughput measurements. The results indicate that Apache Kafka offers the most balanced performance when reliability and scalability are prioritized, while demonstrating that protocol selection should consider both performance metrics and architectural characteristics. The study confirms the applicability of the AHP–TOPSIS methodology as a structured framework for communication protocol selection in M2M sensor data systems.
