Institutional Repository of Academic Research University of Macedonia
Not a member yet
2215 research outputs found
Sort by
Enhancing Security in Federated Learning: Detection of Synchronized Data Poisoning Attacks
No full textFederated learning systems face critical security risks from data poisoning attacks, where malicious clients manipulate training data to compromise model integrity. Traditional detection methods focus on isolating clients that frequently deviate from the average weight update across training rounds. Building upon this concept, this paper introduces an advanced detection strategy that identifies malicious clients through the analysis of similarities in their updates rather than deviations from the average. Our method computes the Euclidean distance between clients’ weight updates vectors over the training rounds. If some clients consistently appear in close proximity to each other, beyond a predefined threshold, they are flagged as potentially malicious. This approach not only refines detection by focusing on synchronization patterns among attackers but also enhances the robustness of the federated model against coordinated data poisoning attacks. We demonstrate the efficacy of our detection method through systematic experiments and discuss optimal hyperparameter tuning strategies, offering a significant step forward in securing federated learning environments.15462211222Artificial Intelligence: Methodology, Systems, and Application
Machine learning techniques in bankruptcy prediction: A systematic literature review
No full textThe main objective of this systematic literature review is to unveil the prevailing trend of employing cutting-edge models for bankruptcy prediction for a period spanning from 2012 to mid-2023. Employing the PRISMA method, we reviewed 207 empirical studies on bankruptcy prediction. Prior extensive research has shown that the integration of more advanced techniques, such as hybrid model, enhances prediction accuracy and robustness, leading to more reliable bankruptcy forecasts. While financial ratios have traditionally played a central role in bankruptcy prediction models, this review places emphasis on the significance of incorporating non-financial ratios. Non-financial ratios capture qualitative and intangible factors such as management competence, corporate governance practices and market reputation. The inclusion of these non-financial ratios, alongside financial ratios in hybrid models, enables a more comprehensive evaluation of a firm's financial health and improves the accuracy of bankruptcy predictions. The review also addresses the challenges and limitations associated with hybrid models and the incorporation of non-financial ratios. Our research shows that there is a current trend toward the development of hybrid models that combine multiple methodologies and variables to improve bankruptcy prediction accuracy. Researchers are actively addressing the challenge of imbalanced datasets in bankruptcy prediction by exploring and developing specialized techniques for handling such data. Moreover, during the evaluation of bankruptcy prediction models, it is essential to consider a range of metrics, including sensitivity and specificity, along with other relevant metrics, to obtain a comprehensive assessment of model performance.25512476
tCOFELET: Conceptual Framework for Team-Centric e-Learning and Training
No full textDespite the advancements in cyber security serious gaming, team-centric approaches have not been explored and the effectiveness of such approaches remains largely untapped. In this light, the main design trends and considerations of multiplayer and collaborative serious games are analyzed, along with weaknesses we identified in the field. Based on this analysis an extension of the Conceptual Framework for eLearning and Training is introduced, named Conceptual Framework for Team-Centric eLearning and Training (tCOFELET). The tCOFELET framework emphasizes the significance of team-centric learning and training and proposes a structured gameplay involving the distinct collaborative phases in planning, performing actions, and reflecting on achievements. tCOFELET integrates COFELET's main elements along with new concepts aiming to create immersive and engaging learning experiences that convey technical knowledge with practical application of skills and development of soft skills such as communication, teamwork, and strategic thinking. Based on the tCOFELET framework a blueprint of a prototype collaborative cybersecurity serious game was elaborated, named mHackLearn. A systematic presentation of mHackLearn's design, analyzed under the prism of the Activity Theory Model for Serious Games (ATMSG), is presented. Finally, a preliminary evaluation of the mHackLearn's game design is performed providing an initial estimation of its potential impact. The findings of the evaluation show that mHackLearn seamlessly integrates several key design considerations of collaborative games, providing promising insights into the tCOFELET's capability to facilitate effective team-centric cybersecurity serious game approaches.12788787889
Exploring the Architectural Composition of Cyber Ranges: A Systematic Review
No full textIn light of the ever-increasing complexity of cyber–physical systems (CPSs) and information technology networking systems (ITNs), cyber ranges (CRs) have emerged as a promising solution by providing theoretical and practical cybersecurity knowledge for participants’ skill improvement toward a safe work environment. This research adds to the extant respective literature, exploring the architectural composition of CRs. It aims to improve the understanding of their design and how they are deployed, expanding skill levels in constructing better CRs. Our research follows the PRISMA methodology guidelines for transparency, which includes a search flow of articles based on specific criteria and quality valuation of selected articles. To extract valuable research datasets, we identify keyword co-occurrences that selected articles are concentrated on. In the context of literature evidence, we identify key attributes and trends, providing details of CRs concerning their architectural composition and underlying infrastructure, along with today’s challenges and future research directions. A total of 102 research articles’ qualitative analyses reveal a lack of adequate architecture examination when CR elements and services interoperate with other CR elements and services participating, leading to gaps that increase the administration burden. We posit that the results of this study can be leveraged as a baseline for future enhancements toward the development of CRs.16723
A Comparison of the Effectiveness of ChatGPT and Co-Pilot for Generating Quality Python Code Solutions
No full textArtificial intelligence (AI) has become increasingly popular in software development to automate tasks and improve efficiency. AI has the potential to help while developing or maintaining software, in the sense that it can produce solutions out of a textual requirement specification, and understand code to provide suggestion on how a new requirement could be implemented. In this paper, we focus on the first scenario. Two AI-powered tools that have the potential to revolutionize the way software is developed are OpenAI’s ChatGPT and GitHub’s Copilot. In this paper, we used LeetCode, a popular platform for technical interview preparation and personal upskilling (self-learning), to evaluate the effectiveness of ChatGPT and Copilot on a set of coding problems, along with ChatGPT’s ability to correct itself when provided with feedback. The analysis of the effectiveness can lead to various conclusions, such as on if these solutions are ready to take over coding roles, and to what extent several parameters (difficulty and quality requirements) influence this result. Solutions have been generated for 60 problems using ChatGPT and Copilot, for the Python programming language. We investigated the performance of the models, the recurrent kinds of errors, and the resulting code quality. The evaluation revealed that ChatGPT and Copilot can be effective tools for generating code solutions for easy problems while both models are prone to syntax and semantic errors. Small improvements are observed for ode quality metrics across iterations, although the improvement pattern is not consistently monotonic, questioning ChatGPT’s awareness of the quality of its own solutions. Nevertheless, the improvement that was found along iterations, highlights the potential of AI and humans, acting as partners, in providing the optimal combination. The two models demonstrate a limited capacity for understanding context. Although AI-powered coding tools driven by large language models have the potential to assist developers in their coding tasks, they should be used with caution and in conjunction with human coding expertise. Developer intervention is necessary not only to debug errors but also to ensure high-quality and optimized code.931012024 IEEE International Conference on Software Analysis, Evolution and Reengineering - Companion (SANER-C
Exploring the Cost Effectiveness of Services in Academic Libraries: A Case Study with the Use of Time-Driven Activity-Based Costing
No full textOver the past decade, the financial crisis has led to reduced government funding for academic libraries in Greece. Now more than ever, it is imperative for library managers to improve their knowledge and understanding of cost behavior, in order to effectively deliver high quality services at decreasing costs. To do so, they need to apply clearly-defined costing methods, such as Time-Driven Activity-Based Costing (TDABC), that allow them to identify the various costs involved in the library processes. In our study, we applied the TDABC method at the medium-sized library of the University of Macedonia (UoM), in Thessaloniki, Greece, to evaluate the costs of the Inter-library Loans (ILL) services. Since the library managers did not adopt a cost allocation method, the cost estimation of the UoM ILL services was rather simplistic and rudimentary. Our research provides empirical evidence of the advantages of TDABC in an academic library setting. Namely, the TDABC method can help library administrators decide how to successfully allocate the available resources and improve the efficiency of the library processes85218720
Software Skills Identification: A Multi-Class Classification on Source Code Using Machine Learning
No full textIn the ever-evolving tech industry, accurately assessing the software skills of developers is critical for effective workforce management. This study presents a machine learning approach to classify software development knowledge through source code analysis, focusing on Java-based technologies. A dataset of several source code files from multiple domains of software development was compiled from public repositories and labeled for classification. The high performance achieved in this study, by applying transfer learning, underlines the suitability of pre-trained CodeBERT models for the classification of software skills. The methodology combined both non-pretrained neural networks and pretrained models to enhance classification accuracy. Results validate the feasibility of using machine learning to identify developers’ programming proficiencies, providing a foundation for sophisticated assessment tools. Future work aims to refine classification by incorporating functional task identification and commit-based analysis for a more comprehensive evaluation of coding skills. This study showcases the transformative potential of machine learning in streamlining developer assessments and advancing software engineering methodologies.6Special Issue 6747
Accounting Outsourcing in Tourism SMEs and Financial Risk Mitigation
No full textThis paper aims to investigate the characteristics of outsourcing in accounting services for tourism SMEs as a choice to mitigate their financial risk. The research was carried out in summer 2022, during tourism recovery from the COVID-19 pandemic crisis, while the findings indicate that the majority of tourism SMEs choose to outsource their accounting services in order to reduce operating costs; to save their funds by exploiting a partner’s information systems; to take advantage of a partner’s accounting knowledge; to achieve greater flexibility in their core activities; and to speed up the processing of the accounting tasks in order to deal with any arising problems and/or difficulties. Furthermore, it is evident that in a constantly changing and complex tax system and a changing economic landscape, accounting outsourcing provides tourism SMEs with advantages such as already established processes, expertise, technology, consulting support, and pathways for dealing with the various accounting issues that may arise.171252
A Multi-model Recurrent Knowledge Graph Embedding for Contextual Recommendations
No full textRecommenders can be improved by exploiting the huge disposal of multi-context data that is now available. Knowledge Graphs (KGs) offer an intuitive way to incorporate this kind of assorted data. This paper introduces a context-aware recommender, based on deriving graph embeddings by learning the representations of appropriate meta-paths mined from a graph database. Our system uses several LSTMs to model the meta-path semantics between a user-item pair, based on the length of the mined path, a Multi-head Attention module as an attention mechanism, along with a pooling and a recommendation layer. Our evaluation shows that our system is on par with state-of-the-art recommenders, while also supporting contextual modeling.14629 LNCS99114Web Engineerin
Fostering Multilingual Deliberation through Generative Artificial Intelligence
No full textDemocracies worldwide face a plethora of challenges, ranging from electoral interference and disinformation to the rising of populism and authoritarianism. It is, hence, imperative to increase participation and broaden access to deliberative processes in order to strengthen democratic institutions and meet public expectations. However, despite the acknowledged importance of language in political deliberation, collaboration, and negotiation, little is known about how multilingualism affects politics and governance. In this context, this study proposes a comprehensive framework that enables multilingual deliberations based on state-of-the-art generative AI technologies. The framework identifies five key offerings namely, “Multilingual and Multicultural Deliberation Design”, “Machine Translation and Interpretation for Citizen Deliberation”, “Multilingual Deliberation Comprehension”, “Online and Face-to-Face Multilingual Deliberation Support”, and “Transparency, Trustworthiness, and Explainability in Citizen Deliberation”. By utilizing Generative AI, the framework intends to address challenges related to cultural diversity and multilingualism that impede successful deliberative democracy. Lastly, a case study is presented that operationalises the framework into a technical solution.3737Proceedings EGOV-CeDEM-ePart conferenc