Scientific articles
Scientific articles
Loss Functions for Time Series Forecasting in Network Security Situation Awareness
Annotation: The choice of the loss function is a key factor in machine learning, especially in training neural networks, as it affects convergence, generalization, and prediction accuracy. In this paper, we compare 20 regression loss functions for training LSTM networks for time series. We also introduce a new loss function, Angle Loss, designed for Network Security Situational Awareness (NSSA). The experiments are based on real cyber alert data from the Warden system. The results, evaluated using the MAE and MASE metrics, show the impact of the choice of loss function on prediction performance.
Doors as Visual Landmarks for Indoor Positioning
Annotation: Indoor localization using a smartphone and the PDR method suffers from an accumulation of errors due to inaccurate sensors and step length estimation. The accuracy can be improved by using Bayesian filtering and additional sources such as maps or Wi-Fi signals. This paper investigates position correction using door detection from camera images using the Grounding DINO model. The detected doors are compared with the map and the position estimate is adjusted accordingly. The experiment showed a reduction in the error from 3.75 m to 1.15 m, while also addressing the anonymization of persons for privacy reasons.
Aplikácia DSA a jeho prienik do ochrany osobných údajov s ohľadom aj na etické princípy
Annotation: Príspevok sa zaoberá vzťahom medzi reguláciami Digital Services Act (DSA) a General Data Protection Regulation (GDPR) v kontexte digitálnych služieb. Analyzuje ich prieniky najmä v oblastiach transparentnosti algoritmov, moderovania obsahu a profilovania používateľov. Pozornosť venuje aj napätiu medzi ochranou základných práv a požiadavkami na bezpečné a spravodlivé online prostredie. Zároveň zdôrazňuje význam etických princípov a potrebu harmonizácie právnych a technologických prístupov pri implementácii DSA.
Communal Level Cybersecurity Incidents in the Slovak Republic
Annotation: The study addresses cybersecurity as a key topic of the digital transformation of society, especially at the level of cities and municipalities. The growing use of digital solutions supports development, but at the same time increases their vulnerability to cyber attacks. The authors analyze 5 cases of attacks in Slovakia and examine the importance of a "bottom-up" approach in building security. The results confirm that local governments play an important role in creating local security policies and responding to incidents. The study thus contributes to empirical knowledge about cybersecurity management at the local level.
Time series dataset for network security situational awareness
- Authors:
- R. Staňa, P. Pekarčík, P. Sokol & V. Bartoš
- Conference/Journal:
- Data in Brief
- Year:
- 2026
Annotation: In the field of Network Security Situational Awareness (NSSA), the lack of high-quality and usable datasets is a big problem. Existing datasets are often outdated, small, or not publicly available. This article presents a new large-scale dataset suitable for neural networks, based on well-documented data. The dataset contains time series of cyber alerts from the Warden system from 2017–2018 and 2023–2024. In total, approximately 3 billion alerts were processed, with data labeling provided by the detection system itself (the so-called silver standard).
Mica Regulatory Framework and Cybersecurity: New Horizons of Consumer Protection in Slovakia
Annotation: As the popularity of crypto-assets increases, so does the number of cyber threats to consumers, which requires a strong legal and institutional framework. The article analyzes consumer protection in the context of the MiCA regulation at both the EU and Slovak levels. It assesses to what extent the new rules and supervision by authorities such as the National Bank of Slovakiacontribute to protection against cyber threats. The results indicate that the legislation brings important protective mechanisms, but also contains shortcomings. These need to be addressed in the future to strengthen consumer protection in the crypto-asset market.
Formal Concept Analysis as a Framework for Cyber Situational Awareness
Annotation: Cyber situational awareness relies on the ability to analyze large numbers of alerts from IDS systems, and identifying hidden patterns and anomalies is challenging. This paper uses Formal Concept Analysis (FCA) to analyze alerts from the dataset CIC-IDS2017 processed using various Snort rules. Based on formal contexts and implication rules, dependencies between rule categories and behavioral patterns are identified. Anomaly events are also detected by using association rules with high confidence. The results show that FCA provides a systematic approach to extracting knowledge from IDS alerts and supports a more effective response to cyber threats.
Legal aspects of artificial intelligence and cybersecurity: data protection and liability in the era of cyber threats
Annotation: In an era of increasing cyberattacks, often supported by artificial intelligence, the protection of personal data is becoming a key challenge. AI enables sophisticated attacks such as phishing and deepfake, increasing the risk of misuse of sensitive data. The article analyses legal aspects, including liability for damage caused by autonomous systems and the need for international regulation. It highlights the importance of a harmonised approach within the EU and a strengthened legal framework. It also highlights the importance of ethics and transparency for the protection of individual rights.