Repository logo

COMPUTER SCIENCE (CN-csci)

Permanent URI for this communityhttps://repo.agh.edu.pl/handle/AGH/102745

...

The main areas of interest of the journal are theoretical aspects of computer science, soft computing, HPC, cloud and distributed processing and simulation, multimedia systems and computer graphics, and natural language processing.

New!   Aktualny numer: 2025 - Vol. 26 - No. 4

Roczniki i numery

Browse

Filters

Advanced Search

Filter by

Settings

Search Results

Now showing 1 - 10 of 556
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    FL-MEC: Federated learning for network traffic classification on the network edge
    (Wydawnictwa AGH, 2025) Paszko, Patryk; Konieczny, Marek; Zieliński, Sławomir; Kwolek, Bartosz
    Nowadays, two technological trends, Federated Learning (FL) and Edge Computing (EC), are increasingly important and influential. FL is a decentralized machine learning strategy that allows learning on distributed data. It primarily allows performing learning operations close to the user, where the data is gathered. This approach belongs to the EC domain, where the main goal is to move computation closer to the end user (e.g., from the centralized cloud). In our work, we apply the FL and EC in the context of network flow classification. We achieved an accuracy of 0.957 with the FL model, compared to 0.924 for the best local model. We achieved these results thanks to the federated averaging performed on neural network layers. To verify our approach, we executed all our experiments on a virtualized environment that emulates existing mid-scale EC network infrastructure, including limitations related to resource constraints on edge nodes.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    Bielik7B v0.1: Polish language model – development, insights, and evaluation
    (Wydawnictwa AGH, 2025) Ociepa, Krzysztof; Flis, Łukasz; Wróbel, Krzysztof; Gwoździej, Adrian; Kinas, Remigiusz
    We introduce Bielik 7B v0.1 – a seven-billion-parameter generative text model for Polish language processing. Trained on curated Polish corpora, this model addresses key challenges in language model development through innovative techniques; these include Weighted Instruction Cross-Entropy Loss (which balances the learning of different instruction types) and Adaptive Learning Rate (which dynamically adjusts the learning rate based on training progress). To evaluate performance, we created the Open PL LLM Leaderboard and Polish MT-Bench – novel frameworks assessing various NLP tasks and conversational abilities. Bielik 7B v0.1 demonstrates significant improvements, achieving a ninepercentage- point increase in its average score compared to Mistral-7B-v0.1 on the RAG Reader task. It also excels in the Polish MT-Bench – particularly in the Reasoning (6.15/10) and Role-playing (7.83/10) categories. This model represents a substantial advancement in Polish language AI, offering a powerful tool for diverse linguistic applications and setting new benchmarks in the field.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    Toward RAM forensics supported by machine-learning methods
    (Wydawnictwa AGH, 2025) Jurczyk, Kamil; Topa, Paweł; Faber, Łukasz
    In this article, we propose an enhancement to the computer forensics technique of using Machine-Learning tools to analyze the contents of RAM in order to extract information that is potentially useful during an investigation. In the specific case presented, the use of the extracted information to generate moreoptimal dictionaries for dictionary cryptanalysis is considered. Increasing user awareness is making cryptanalysis of passwords increasingly difficult for law enforcement. Long and complex passwords are impossible to crack – even when high-performance computing platforms are available. A sensible method of optimization is to look for hints to use a dictionary that contains text phrases more likely to be used in the specific case under attack. Such a hint could be an analysis of RAM taken from a suspect computer. Machine-learning methods can significantly facilitate this task. In this article, we also explore the effectiveness of such an approach and its usefulness in practical applications. We also consider applications of the proposed approach for other purposes, such as OSINT.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    The benefits of testing software in se research: lessons learned from two phd projects
    (Wydawnictwa AGH, 2025) Novak, Matija; Mijač, Marko
    Software engineering (SE) research often involves creating software – either as a primary research output (e.g., in design science research) or as a supporting tool for the traditional research process. Ensuring software quality is essential, as it influences both the research process and the credibility of findings. Integrating software-testing methods into SE research can streamline efforts by addressing the goals of both research and development processes simultaneously. This paper highlights the advantages of incorporating software testing in SE research – particularly for research evaluation. Through qualitative analysis of software artifacts and insights from two PhD projects, we present ten lessons learned. These experiences demonstrate that, when effectively integrated, software
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    A parallel approach for metaheuristics solving the labs problem using CPU and GPU
    (Wydawnictwa AGH, 2025) Żurek, Dominik; Piętak, Kamil; Pietroń, Marcin; Kisiel-Dorohinicki, Marek
    This paper contributes to solving the low autocorrelation binary sequence (LABS) problem that remains an open hard-optimization problem with many applications. The current direction of research is focused on developing algorithms dedicated to parallel architectures such as GPGPU or multi-core CPUs. The paper follows this direction and proposes new heuristics developed from the steepest-descent local search algorithm that extends the notion of a neighborhood of a given sequence. The introduced algorithms utilize the parallel nature of multicore CPUs and provide an effective method for solving the LABS problem. The efficiency levels of SDSL and the new algorithm are presented; to ensure an effective comparison, they were both implemented in the same manner. The comparison shows that exploring the larger neighborhood improves the efficiency of the search method.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    Optchain: an advanced optimization method for enhancing IoT Data security via blockchain
    (Wydawnictwa AGH, 2025) Kokate, Shatakshi; Shrawankar, Urmila
    The increased use of IoT devices in various domains generates abundant data traffic. Securing this data during its transfer and storage is essential. Blockchain is now a trending technology to provide security to the data; however, it is observed that blockchain performs poorly while managing large volume data. To mitigate this issue, an advanced Optchain method to reduce the data size before submitting it to the blockchain network is discussed in this paper. This Optchain method optimizes IoT-generated data using data-classification and compression techniques. The classification of data as relevant or irrelevant is based on predefined thresholds of critical healthcare parameters. Subsequently, the Optchain method employs the Z-standard algorithm for compressing only the relevant data, ensuring efficient storage and faster blockchain transactions. Simulation results using the iFogSim simulator and Ethereum blockchain demonstrated improved storage costs and computational times compared to traditional methods.
  • thumbnail.journal.alt
    Item type:Journal Issue,
    Computer Science
    2025 - Vol. 26 - No. 4
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    PSO-WESRGAN: a novel document image super resolution
    (Wydawnictwa AGH, 2025) Kezzoula, Zakia; Gaceb, Djamel
    Abstract One of the major challenges of document images that can hinder readability and the analysis of information is low resolution; this is typically caused by low-pixel density scanning or excessive compression to save storage space. This results in a loss of fine detail in images, making it difficult to detect critical information. To solve these problems, super-resolution techniques are used. These techniques improve image quality by increasing the resolution while maintaining the fine detail. PSO-WESRGAN is an innovative method that combines wavelet processing, deep-transfer learning, and particle swarm optimization (PSO). Wavelet processing analyzes image detail at diverse scales and orientations, while transfer-based deep-learning advantages pre-trained models on vast image data sets. By integrating PSO, the efficiency of the method is enhanced through the optimal exploration of the solution space to identify the best parameters for the super-resolution model. The experimental results show the effectiveness of this method and open up prospects for future improvements in the super-resolution of document images.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    Can artificial intelligence predict a tsunami?
    (Wydawnictwa AGH, 2026) Wójcik, Daria; Niewiadomska, Alicja; Paszyński, Maciej
    In this article, we build a model for tsunami simulation based on physicsinformed neural networks and the Finite Difference Method. We then check how the numerical results obtained using these two methods differ from each other. Assuming that the Finite Difference Method gives accurate results, we estimate the error resulting from the use of physics-informed neural networks. We compare this estimate with surveys conducted among computer science students in order to assess the level of public trust among specialists in the numerical results obtained using artificial intelligence tools. In particular, we assess how reliable tsunami predictions obtained using physics-informed neural networks are and what the public perception of the reliability of such predictions is.
  • thumbnail.journal.alt
    Item type:Article, Access status: Open Access ,
    Attention-based multiple-representation method for fingerprint-presentation-attack detection
    (Wydawnictwa AGH, 2025) Uttam, Atul Kumar; Agrawal, Rohit; Jalal, Anand Singh
    Fingerprint biometrics are one of the most common authentication mechanisms; however, such systems are often compromised by presentation attacks that are made by presentation-attack instruments. Most fingerprint-presentationattack- detection approaches show poor performance due to the large variations in the presentation-attack instruments and the limited feature representation of the input fingerprint. Therefore, this article proposes a hybrid model of shallow and deep features with multiple representations of input fingerprints. To obtain these shallow and deep features, we first enhanced the texture of the input fingerprint through a novel median adaptive local binary pattern filter and an existing binarized statistical image feature. After this, the input fingerprint image and two textured enhanced images are concatenated along with the channel dimension for multiple representations. Finally, an extended ResNeXt architecture with channel and spatial attention (EResNeXt) was used for relevant feature extraction and presentation attack detection. EResNeXt was evaluated in the LivDet-2015 and LivDet-2017 data sets, and ACEs (average classification errors) were obtained at 0.94 and 0.49, respectively.