Gait, an individual's unique walking style, serves as an effective biometric tool for surveillance. Unlike fingerprints or iris scans, gait is observable from a distance without the subject's awareness, making it ideal for security applications. CNNs struggle with video variability, affecting gait recognition. This study introduces GaitDeep, a spatial-temporal refinement using a deep dense network. It integrates attention-enhanced spatial extraction with a two-directional LSTM-based temporal module to prioritize key segments. Evaluated on the OU-ISIR, OU-MVLP, and CASIA-B datasets, GaitDeep achieves accuracies of 95.1%, 0.96%, and 98.10%, respectively, outperforming state-of-the-art methods and establishing a new benchmark for gait recognition.

The proliferation of real-time, infinite data streams necessitates efficient online learning approaches. Hoeffding Trees (HT), which extend traditional decision trees using the Hoeffding bound, offer robust stream classification but face high computational costs. While the Green Accelerated Hoeffding Tree (GAHT) addresses energy efficiency concerns, its prediction accuracy can be improved by addressing its inherent limitations in combining Hoeffding bounds with information gain metrics for incrementally growing the tree. This study successfully develops enhanced GAHT variants through optimized Hoeffding bound stability and node splitting mechanisms. Our empirical evaluation demonstrates that the usage of these new variants improves predictive performance over the state-of-the-art GAHT, without compromising its energy efficiency.

This study presents a feature-level analysis of text complexity using large language models (LLMs) in a two-phase design. Phase I operationalized six core features - lexical diversity, density, syntactic complexity, coherence, named entities, and readability - achieving Spearman correlations of 0.55-0.60 across domains. Phase II employed indirect prompting to surface additional qualitative dimensions (e.g., inferential load, rhetorical structure), yielding a mean correlation of 0.42 and revealing that the six features account for 40% of complexity variance. Domain dependencies were limited to named entities and lexical diversity. We propose a hybrid model combining normalization, root-based synergies, and newly quantified metrics with domain-tuned formulae for improved prediction.

Camera position is essential for many applications, such as monitoring, tracking, and recognizing individuals. This study proposed an integrated design that combines recurrent neural networks (RNNs) and a loss function modification approach to improve the accuracy of indoor camera location. RNNs enable the system to generate accurate estimations based on previous information by extracting temporal dependencies and patterns from the camera information. We optimized the loss function to enhance the indoor camera position's overall performance and convergence speed. This combination technique allows the proposed method to considerably increase the accuracy of camera location prediction in indoor conditions. We validated the effectiveness of the proposed approach and demonstrated its improved accuracy and robustness through extensive evaluation of many indoor datasets. The results show that our combined approach outperforms existing methods and has enormous potential for real-world applications in indoor activity recognition, navigation optimization systems, and safety surveillance.

The study examines possible solutions to resolve problems that occur in the process of operationalizing systems that predict the financial services that customers will terminate using machine learning and business intelligence approaches. The scope of the study consists of defining the problem, collecting and integrating the data, training the models, evaluating and validating the outputs, and making them ready for use in the production environment. In addition, intuition about infrastructure, architecture, processes, technologies, and other artifacts used during the study is included. The data manipulation and pre-processing framework proposed in this study is applicable to both real and synthetic banking data. To implement each step in detail, an improved version of an auxiliary study was used. A study has been carried out in a financial institution in Turkey, chosen as an auxiliary, in which customers who are likely to cancel credit cards are determined by machine learning. The problems, findings, and results are examined in detail. The framework used in this study is believed to be used not only in the integration of credit card product churn detection systems but also in the integration of other systems that use machine learning and deep learning.

This paper describes the steps through which the authors passed during the process of digitization of manually written mathematical texts with formulas and figures. Some difficulties met are also discussed. Our project highlighted the challenges associated with working with handwritten, non-homogeneous texts stored on outdated records, but it also demonstrated the effectiveness of combining modern technology with traditional manual methods.

The article includes the modified deep first search algorithm (DFS) that allows, at a single traversal of a graph, to check its connectivity/biconnectivity, highlight the cut vertices, and build the spanning tree, the biconnected components, and the fundamental set of cycles. The proposed algorithm was implemented and tested in a functional style using \textsc{Common LISP} language. \footnote{The project SIBIA 011301 has supported a part of this research.

This paper offers a gentle introduction into the realm of monotone span programs and their connection with linear secret sharing schemes and attribute-based encryption while emphasizing the cryptographic importance of finding efficient MSPs for representing complex access structures. We provide a proof that there is no ideal LSSS for Boolean circuits, thus tackling the open problem of finding LSSSes of non-exponential size for Boolean circuits. Moreover, we present an application of our proof to graph access structures and a backtracking approach to finding efficient MSPs for given access structures.

Our paper focuses on Linked Data Lineage as the collection of modeled organizational data, enabling continuous integration and modern governance in Business-to-Business (B2B) sharing. We first explore approaches organizations adopt to enhance B2B data integration. Next, we highlight key B2B and Business-to-Government (B2G) linked data use cases and strategies for governance frameworks. We review software tools supporting data governance and propose a systemic method for defining organizational integration flows. This is demonstrated through a system enabling businesses to design linked data pipelines, leveraging Semantic Web technologies for continuous integration and governance. Insights and recommendations for improving data sharing processes are provided.

The present research focuses on the security of Macedonian websites. It involves the analysis of HTTP Security header responses for 756 websites in the country, of which 246 are the most popular. This analysis is conducted across 13 different categories of websites, including government bodies and institutions, public institutions and enterprises, educational, commercial, news and media, entertainment, sports, etc. We intend to create a comprehensive security profile for the country's websites, which will help raise their overall security level. It is critical to understand and implement proper HTTP security headers to prevent or limit the dangers that can cause website attacks such as Denial of Service (DoS), Cross-Site Scripting (XSS), Cross-Site Request Forgery (CSRF), SQL Injection, clickjacking, etc. Our analysis was performed with the help of the Mozilla Observatory tool. We have discovered a significant lack of implementation and/or misconfiguration of HTTP security headers in all categories. Almost half of the websites (n=375; 49.60\%) have an F grade, while more than a quarter of all websites (n=214; 28.31\%) have a minimal security score of 0.