XL International Conference Infotech 2025 Proceedings

XL International Conference INFOTECH 2025 PROCEEDINGS Edited by Vladan Pantović Dušan Starčević Aranđelovac, 4 – 5, June, 2025 XL International Conference INFOTECH 2025 Aranđelovac, 4 – 5, June, 2025 Organizer ASIT - Association for Computing, Informatics, Telecommunications and New Media of Serbia Co-organizers of the scientific part of the conference: • Faculty of Project and Innovation Management prof. dr Petar Jovanović, Belgrade • Faculty of Information Technology and Engineering, Belgrade • Faculty of Business Economy and Entrepreneurship, Belgrade • Laboratory for multimedia communication, FON, Belgrade • University “Bijeljina”, Bijeljina, Bosnia and Herzegovina • “Dositej” Faculty of Economics and Informatics, Belgrade INFOTECH 2025 Proceedings Edited by Vladan Pantović & Dušan Starčević Publisher: ASIT, Belgrade, Nikola Mirković, President Co-Publisher: Faculty of Project and Innovation Management prof. dr Petar Jovanović, Belgrade Cover Design: Vladimir Jablanov / Digital printing: GRAFOPAN doo / Circulation: 200 ISBN-978-86-900491-3-4 CIP - Каталогизација у публикацији Народна библиотека Србије, Београд 004(082) 007:004(075.8)(082) INTERNATIONAL Conference INFOTECH (40 ; 2025 ; Aranđelovac) Proceedings / XL International Conference INFOTECH 2025, Aranđelovac, 4 – 5, June, 2025 ; edited by Vladan Pantović, Dušan Starčević ; [organizer] ASIT [i. e.] Association for Computing, Informatics, Telecommunications and New Media of Serbia ; [co-organizers Faculty of Project and Innovation Management ... [et al.]]. - Belgrade : ASIT : Faculty of Project and Innovation Management, 2025 (Beograd : Grafopan). - 138 str. : ilustr. ; 30 cm Tiraž 200. - Str. 5: Preface / Nikola Mirković, Vladan Pantović. - Bibliografija uz svaki rad. ISBN 978-86-900491-3-4 (ASIT) 1. Pantović, Vladan, 1961- [urednik] [autor dodatnog teksta] 2. Starčević, Dušan, 1949- [urednik] a) Информациона технологија -- Зборници b) Информациони системи -- Зборници COBISS.SR-ID 169843977 XL International Conference INFOTECH 2025 Proceedings 3 Table of Contents PREFACE ………………………………………………………………………………………………………………………………….. 5 PROGRAM COMMITTEE …………………………………………………………………………………………………………… 6 INVITED KEYNOTE LECTURE: Trust in the Age of AI: Decoding the Future of Data-Driven Insights Nikola Vojtek, Vladan Pantović ……….…. 9 1. Artificial Intelligence ………………………………………………………………………………………………………….. 13 • Application of Artificial Intelligence Technologies in Autonomous Vehicles – Advantages and Challenges, Radoslav Raković…………………………………………………………………………………………. 15 • Code Generators in the Age of Large Language Models, Mladen Opačić, Nikola Dimitrijević, Nemanja Zdravković…………………………….…………………………………………………………………………. 21 • AI’s Influence on Organizational Culture in Software SMEs in Southeast Europe: Challenges and Metrics, Aleksandar Milinčić…………………………………………………………………………………… 27 • Driving Innovation through Intelligent Automation, Dragan Metikoš………………………………. 33 • Tackling Data Scarcity in AI: A Comparative Analysis of Data Augmentation and Synthetic Data Generation, Nikola Vojtek, Bojan Smudja ……………………………………………………………….. 37 • Using Artificial Intelligence in Information Technology Audits, Dragan Jovičić, Kristijan Lazić, Vladan Pantović, Marina Jovanović-Milenković, Ivan Vulić …………………………………………..…. 43 • Organizational Resilience and Competitiveness with ISO/IEC 42001: A Framework for AI Data Supervision, Nikola Vojtek, Vladan Pantović……………………………………………………………. 49 • Integration of Artificial Intelligence Tools in Peer-Review: Transparency, Efficiency and Reliability of Hybrid Model, Dragorad Milovanović, Zoran Jovičić, Siniša Ristić……………….. 53 • Leveraging KNIME for AI Model Design in IoT and Edge Computing Scenarios, Petar Prvulović, Nemanja Radosavljević, Đorđe Babić, Dušan Vujošević……………………………………. 57 • Internal Audit in K–12 Education in the Age of Artificial Intelligence: Challenges, Potentials, and Recommendations, Sonja Djukic Popović, Stefan Popović, Dražen Jovanović …………… 63 • Human and Artificial Intelligence in Team Work, Vesna Buha, Rada Lečić, Mirjana Dejanović ……………………………………………………………………………………………………………………….. 69 2. Information Security ………………………………………………………………………………………………………….. 73 • AI's Role in Cybersecurity Threats and Defenses - Examples of Concrete Solutions, Dragan Pleskonjić, Luka Tica, Vladimir Jelić, Dušan Todorović, Anthony English ……….…………………. 75 • Sirius E-Learning Platform for Creating and Delivering Accredited Online Seminars - IT Security in Schools, Vladan Stevanović ...………………………………………………………..………………. 81 • Risk assessment of the Information System by Applying the AHP Method, Branko Vujatović, Sanja Klajić, Darko Grubač, Marija Vujatovć, Nenad Stojanović ………………………….………….. 85 XL International Conference INFOTECH 2025 Proceedings 4 3. Information Technology and Applications ………………………………………………………………………….. 91 • Exploring the Application of Blockchain and Smart Contracts in Construction Progress Payments, Paolo Eugenio Demagistris, Filippo Maria Ottaviani .…………………………..………… 93 • Tool for Literature Discovery and Analysis, Nikola Vojičić, Davor Sekulski, Vladimir Urošević ………………………………………………………………………………………………………………………………………… 99 • Study on 5G-Advanced FinIoT Platfom in Emerging Financial Use Cases, Carol Edrich, Dragorad Milovanović, Drago Indjić …………………………………………………………………………..…. 105 • Process of Migrating Single Page Applications to Meta Frameworks Like Nuxt, Petar Kresoja, Marko Šarac, Mladen Veinović ……………………………………………………………………………………… 109 • Modern Approaches to Water System Security: Integration Scada System and Blockchain, Irena Tasić, Srđan Tasić …………………………………………………………………………………………………. 113 4. Management and Information Systems ……………………………………………………………………………. 119 • Assessment of the use of information and communication technology on the example of an agricultural farm, Miroslav Nedeljković, Slađana Vujičić, Cvijetin Živanović …………………… 121 • Models of Management Organizational Changes, Biljana Ilić, Slavica Anđelić, Sanja Stojanović …………………………………………………………………………………………………………………….. 125 • E-HRM Like Potential of Business, Julija Avakumović, Jelena Avakumović ……………………… 133 WORKSHOP: Implementation of the ISO/IEC 42001 Standard for AI Management System ……. 137 AUTHOR INDEX ……………………………….…………………………………..………………………………………………. 138 XL International Conference INFOTECH 2025 Proceedings 5 P R E F A C E This year marks the 40th edition of INFOTECH, a regular annual international scientific and professional conference in the field of the development and application of information technologies. INFOTECH has always followed trends in the development of information and communication technologies, and this year is no different; the obvious dominant theme, addressed in more than half of the papers, is the application of artificial intelligence (AI). Due to the importance of this current topic, a special AI Panel, AI Workshop and AI Keynote were organized. A total of 23 accepted papers by 58 authors are published in the Proceedings. They are classified according to their subject matter into four sections: Artificial Intelligence, Information Security, Information Technology and Applications, and Management and Information Systems. The tradition continues, and this year INFOTECH also featured authors from abroad (Italy, Canada, United Kingdom, Germany and Bosnia and Herzegovina). ASIT - the Association for Computing, Informatics, Telecommunications and New Media of Serbia, the organizer of the conference, on the occasion of the important jubilee, published, in cooperation with the Co-Publisher the “Faculty of Project and Innovation Management prof. dr Petar Jovanović“, two additional special editions edited by prof. dr Vladan Pantović and prof. emeritus dr Dušan Starčević: • INFOTECH 2020 – 2024 Selected Papers • INFOTECH 1995 – 2000 Proceedings: Digital version the classic printed proceedings We would like to thank everyone who actively participated in the preparation of the INFOTECH 2025 conference, and we expect good cooperation in the coming years as well. Chairman of the Organizing Board Chairman of the Scientific Program Board Nikola Mirković Prof. dr Vladan Pantović XL International Conference INFOTECH 2025 Proceedings 99 ENHANCING LITERATURE REVIEWS A TOOL FOR DISCOVERY AND ANALYSIS Nikola Vojičić, Belit d.o.o. (Ltd.), nikola.v[email protected].rs Davor Sekulski, Belit d.o.o. (Ltd.), [email protected].rs Vladimir Urošević, Belit d.o.o. (Ltd.), vladimir.uro[email protected] Abstract: This paper introduces a custom-built Literature Review Tool desinged to discover, extract, and rank relevant content from a diverse range of online sources using advanced web scraping and contextual full-text search techniques. The tool enhances the efficiency and accuracy of literature reviews by enabling comprehensive analysis of both scientific publications and grey literature, including blogs, informal reports, and government documents. The paper presents an overview of the tool’s core functionalities, including source discovery, article extraction, and content analysis. Keywords: Web scraping and content extraction, full-text search, relevance scoring and highlighting, literature analysis, grey literature. 1. INTRODUCTION The surveying, discovery, and systematization of relevant grey literature materials “produced on all levels of government, academics, business and industry in print and electronic formats, […] which is not controlled by commercial publishers” [1] – such as technical reports, project reports, working papers, white papers, government documents, and other forms – is highly significant [2] for mapping the state of the art with a more balanced view [3], particularly in applied research and innovation, yet it is largely overlooked or unsupported by major academic search engines and structured content repositories, such as Scopus, Lens, ScienceDirect, and Web of Science. In the absence of efficient tools or structured registries, literature reviews of this kind often rely on either manual web searching or interaction with LLM-based interfaces (such as OpenAI ChatGPT [4]). However, these approaches remain highly unreliable for tasks that demand deterministic responses and fully accurate, complete results, due to issues such as hallucinations or confabulations [5] and inherent biases [6]. Manual web searches are effective for relatively simple tasks, such as gathering information on straightforward, narrowly focused topics or addressing specific problems with widely accepted answers and minimal conflicting viewpoints. The limitations of manual searching become evident when conducting extensive and systematic grey literature reviews, especially those that require diverse sources, multiple perspectives, and cover different time periods. Manual searching in such cases is time-consuming and labor-intensive, often necessitating a well-coordinated team, thorough planning, and systematization of results. Even when relevant documents are found, the outcomes of manual searches often fail to justify the time and effort invested. This is especially true when results, despite being well-documented and structured, are not stored in databases that support complex evaluations, aggregations, and analyses. While ad-hoc scripts may assist with some preliminary tasks, they only partially address the challenge of assembling a relevant corpus. Significant additional effort is still required to thoroughly read through these documents and select the most relevant ones. Among all these activities, only two truly require human effort: defining search queries and reading the discovered articles. Everything else can be automated, including ranked searches across the internet for websites and within websites for articles, scraping article content, storing data in structured, queryable database, and visualizing, filtering, aggregating scores, and ordering the data. Even the manual tasks can be optimized: search queries can be prioritized based on the scores of retrieved results, and reading can be minimized by using contextual full-text search to highlight only the most relevant paragraphs and keywords within the extracted article contents. This article describes a tool for automated literature discovery and review designed to address key challenges in the field. Developed by a dedicated project team, it is actively used in the ongoing FishEUTrust project to explore, discover, and systematize grey literature on several of its core topics and research questions, such as barriers to and drivers of seafood purchasing and consumption behaviour. The tool’s development was motivated by the clear absence of any suitable existing solutions at the time (mid-2023). XL International Conference INFOTECH 2025 Proceedings 100 2. MAIN FEATURES AND WORKFLOW The Literature Review Tool is designed to overcome the challenges of manual internet article searches by automating the discovery of relevant articles and sections, while also supporting user collaboration during content analysis. The workflow consists of four key stages: 1. Website search which involves searching websites as sources of specific articles. This step is optional, as users can manually provide a set of websites if they already know where to search. 2. Article search on specific websites involves searching for articles within the subset of websites identified in step 1, websites manually entered by the user, or both. 3. Paragraph search within articles takes place after step 2, once articles have been collected and the search is complete. The tool downloads content from the top 4,000 ranked articles, extracts and indexes their content, and enables full-text search on that data. 4. Tagging websites and articles with statuses such as relevant, potentially relevant, or not relevant to helps with collaboration among multiple users. This tagging can be done at any stage in the workflow. Figure 1. Literature Review Tool workflow Figure 1 illustrates the possible steps in this workflow, detailing the inputs and outputs for each stage. Discovering websites involves website searching and analysis. The article search is performed on the identified relevant websites to collect material for content extraction. Paragraph search locates relevant sections within documents, reducing the need for reading the full text. 2.1 Discovering Sites Searching the entire internet for articles on a specific topic is challenging. The vast amount of information often causes search engines to favor higher-ranked sites optimized for search, which may not always be the most relevant. This can narrow the diversity of results. To address this issue, it is necessary to restrict the search to a predefined set of relevant websites. In some cases, users may already know the specific websites to target for article searches. In other cases, additional websites need to be identified to expand the source pool. This tool conducts website searches by scanning the internet and scraping context paths from the first five pages of search results. Users provide search queries, and each search result is stored as a triplet consisting of site, query, and page. For analysis and visualization, the flat rows are grouped by site, with scores aggregated by summing individual contributions and queries aggregated into sets. Each appearance of a site on a page contributes 10 𝑝𝑎𝑔𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 (1) to the total score of that site. For example, if a site appears four times on the first page, its score will be 40, and if it appears once on the second page, it will add 5 to the score. Figure 2. Left column of the site discoverer page with websites and scores Hovering over the score number, as shown in Figure 2, opens a dropdown list displaying all the queries for which the site appears. Sites can be marked with the following statuses: “rejected” (red), “uncertain” (yellow), and XL International Conference INFOTECH 2025 Proceedings 101 “accepted” (green). Users can also add custom statuses – for example, Figure 2 shows an additional status, “academic” (pink), used for websites containing academic literature. Once a site is marked as “accepted,” it is automatically queued for article search (see the next chapter), while other statuses remain descriptive and searchable only. Figure 3. Right column of the site discoverer page with queries and scores Since this process involves searching across the entire internet, queries should be longer and more specific to narrow the results. The query score helps guide users in refining their queries. For each query, the score is calculated as the sum of 1 𝑠𝑖𝑡𝑒 𝑟𝑎𝑛𝑘 (2) for each site found by that query, where the site with the highest score is assigned rank 1 and subsequent sites receive incrementally higher ranks. This sum is then multiplied by 100. Queries can be deleted, which also removes all associated results. This action cannot be undone, although users can manually re-enter the same query if needed. Checking the boxes for specific queries filters the websites listed in the left column, showing only those with scores that include the selected queries (see Figure 3). Additionally, users can perform a fuzzy search by site or combine these filters. 2.2 Discovering Articles After optionally searching for and analyzing websites, and marking some as accepted, the next step is to search for articles. If websites were not previously identified through a search, they must be manually entered, since both websites and queries are required inputs for the article search. This search then looks for articles matching the specified queries on the designated websites. For the initial search, both a site and a query are required. Subsequent searches allow entering either or both. When a new site is added, all previous queries will be searched on that site. When a new query is added, it will be searched across all previously specified sites. If both a new site and a new query are provided, the new query will be searched across all existing sites, and all previous queries will be searched on the new site. Searches for existing site-query pairs are not repeated. To refresh results, users must remove and re-enter existing entries. This approach ensures consistent scoring and relevant results. The article scraper extracts article URIs and titles from the top three pages of search results. Each site-query pair triggers three scrapes, one for each page. Limiting extraction to the first three pages has proven sufficient, as the search is focused on specific websites rather than the entire internet, resulting in less bias compared to the site search phase. The outcome of each search contains: site, query, page, scraped article URI and title. These flat rows are grouped by article, scores are aggregated by summing individual contributions, while queries and sites are combined into sets. Each appearance of a URI on a page contributes 10 𝑝𝑎𝑔𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 (3) to the article's total score. For example, an article that appears twice on the first page will have a score of 20, and a score of 5 if it appears once on the second page. Figure 4. Left column of the article discoverer page with articles and scores Hovering over the score number opens a dropdown listing all queries for which the article was found. Like websites, articles can be marked with statuses such as “rejected”