Informatics (Id)
Analysis of digital evidence using machine learning methods
Digital forensic analysis has become an essential part of responding to cybersecurity incidents as well as part of cybercrime investigation. An important phase of forensic investigation is the analysis of digital evidence itself. Within this phase, it is necessary to extract forensic artefacts, determine their relevance, value for the case, as well as relationships between them. The purpose of this phase is to confirm, resp. reject the forensic hypotheses established in the early stages of the forensic investigation. The aim of this work is to analyze the possibilities of using machine learning methods in the analysis of digital tracks with respect to the complexity, volume, and heterogeneity of forensic artefacts. At the same time, the aim is to propose a method of selection for the case of relevant forensic artefacts, to find a relationship between them as well as to verify the forensic hypothesis itself.
(1) Analyze the possibilities of using machine learning methods to analyze digital traces considering forensic artifacts' complexity, quantity, and heterogeneity. (2) To propose a method of selecting relevant forensic artifacts and finding their relationship. (3) To propose a way of verifying the forensic hypothesis itself.
(1) Hall, Stuart W., Amin Sakzad, and Kim‐Kwang Raymond Choo. "Explainable artificial intelligence for digital forensics." Wiley Interdisciplinary Reviews: Forensic Science 4.2 (2022): e1434. (2) Mohammad, Rami Mustafa A., and Mohammed Alqahtani. "A comparison of machine learning techniques for file system forensics analysis." Journal of Information Security and Applications 46 (2019): 53-61. (3) Tallón-Ballesteros, Antonio J., and José C. Riquelme. "Data mining methods applied to a digital forensics task for supervised machine learning." Computational intelligence in digital forensics: forensic investigation and applications (2014): 413-428. (4) Du, Xiaoyu, et al. "SoK: Exploring the state of the art and the future potential of artificial intelligence in digital forensic investigation." Proceedings of the 15th International Conference on Availability, Reliability and Security. 2020.
doc. RNDr. JUDr. Pavol Sokol, PhD.
Informatics (Id)
Analysis of judicial decisions using artificial intelligence
The application of artificial intelligence in the decision-making of courts is a subject of the current interest of the European Union, which is declared in several official documents. The decision-making activity performed by the competent court requires knowledge of legal principles, general principles of law and understanding of the legal text. It is a challenge for the application of different methods of artificial intelligence. An important part of the whole automation process is the extraction of knowledge from the texts of court decisions and their subsequent automated analysis. The aim of the work will be to explore the possibilities of extracting the structured attributes of court decisions, which are present in these decisions mostly in the form of free text in natural language. Also, the aim will be to design a model for the extraction of these attributes and its implementation within the system for court decisions searching.
The application of artificial intelligence in the decision-making of courts is a subject of the current interest of the European Union, which is declared in several official documents. The decision-making activity performed by the competent court requires knowledge of legal principles, general principles of law and understanding of the legal text. It is a challenge for the application of different methods of artificial intelligence. An important part of the whole automation process is the extraction of knowledge from the texts of court decisions and their subsequent automated analysis. The aim of the work will be to explore the possibilities of extracting the structured attributes of court decisions, which are present in these decisions mostly in the form of free text in natural language. Also, the aim will be to design a model for the extraction of these attributes and its implementation within the system for court decisions searching.
prof. RNDr. Stanislav Krajči, PhD.
Informatics (Id)
Analysis of structured data using machine learning methods
Machine learning methods and algorithms enable various forms of innovation in education, healthcare, industrial digitalization, and other areas of our life. Machine learning applications thus have significant potential as they can help improve people's lives. The dissertation aims to propose machine learning methods and algorithms for working with structured data and to describe their principles and properties. The work aims to apply supervised or unsupervised learning methods in machine learning in various application domains.
Machine learning methods and algorithms enable various forms of innovation in education, healthcare, industrial digitalization, and other areas of our life. Machine learning applications thus have significant potential as they can help improve people's lives. The dissertation aims to propose machine learning methods and algorithms for working with structured data and to describe their principles and properties. The work aims to apply supervised or unsupervised learning methods in machine learning in various application domains.
prof. RNDr. Stanislav Krajči, PhD.
doc. RNDr. Ľubomír Antoni, PhD.
Informatics (IdAj)
Anomaly detection using machine learning methods
Data analysis solutions are applied in various areas of technical, natural, human and economic sciences. Machine learning is a sub-area of artificial intelligence that deals with machine learning methods and algorithms based on input data in a defined solution space. The aim of the dissertation thesis is to design and application of algorithms and methods of machine learning in case studies of anomaly detection and to compare the performance of the proposed solution with other available studies.
The aim of the dissertation thesis is to design and application of algorithms and methods of machine learning in case studies of anomaly detection and to compare the performance of the proposed solution with other available studies.
1. Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM computing surveys (CSUR), 41(3), 1-58. 2. Pang, G., Shen, C., Cao, L., & Hengel, A. V. D. (2021). Deep learning for anomaly detection: A review. ACM computing surveys (CSUR), 54(2), 1-38. 3. Russell, S., & Norvig, P. (2021). Artificial Intelligence: a modern approach, 4th US ed. University of California, Berkeley.
doc. RNDr. Ľubomír Antoni, PhD.
prof. RNDr. Gabriel Semanišin, PhD.
Informatics (IdAj)
Brain-training games for spatial hearing
Solutions designed to enhance auditory processing when hearing thresholds are within normal limits are very limited and none are as recognized or as widely available as are hearing aids and cochlear implants. The project aims to contribute to the development of novel procedures to rehabilitate auditory processing deficits (APD) by developing a brain training game based on modern auditory neuroscience and the results of the EU Horizon Europe SAV grant. The development of auditory brain training game will be in collaboration with Northeastern University Brain Game Center and Oregon Health State University. The main goal of the games is to develop and test rehabilitative techniques that restore auditory function for those who perform poorly on tests of APD by training various aspects of auditory processing.
Klingel M, Laback B, Kopco N (2021) Reweighting of Binaural Localization Cues Induced by Lateralization Training. Journal of the Association for Research in Otolaryngology, 22, 551–566, https://doi.org/10.1007/s10162-021-00800-8. Spisak O, Klingel M, Loksa P, Sebena R, Laback B, Kopco N (2021) “Spectral and binaural cue reweighting for sound localization in real and virtual environments,” 2nd Joint Conference on Binaural and Spatial Hearing, 7-8 October 2021.
doc. Ing. Norbert Kopčo, PhD.
Informatics (Id)
Brain-training games for spatial hearing
Solutions designed to enhance auditory processing when hearing thresholds are within normal limits are very limited and none are as recognized or as widely available as are hearing aids and cochlear implants. The project aims to contribute to the development of novel procedures to rehabilitate auditory processing deficits (APD) by developing a brain training game based on modern auditory neuroscience and the results of the EU Horizon Europe SAV grant. The development of auditory brain training game will be in collaboration with Northeastern University Brain Game Center and Oregon Health State University. The main goal of the games is to develop and test rehabilitative techniques that restore auditory function for those who perform poorly on tests of APD by training various aspects of auditory processing.
Klingel M, Laback B, Kopco N (2021) Reweighting of Binaural Localization Cues Induced by Lateralization Training. Journal of the Association for Research in Otolaryngology, 22, 551–566, https://doi.org/10.1007/s10162-021-00800-8. Spisak O, Klingel M, Loksa P, Sebena R, Laback B, Kopco N (2021) “Spectral and binaural cue reweighting for sound localization in real and virtual environments,” 2nd Joint Conference on Binaural and Spatial Hearing, 7-8 October 2021.
doc. Ing. Norbert Kopčo, PhD.
Informatics (IdAj)
Cross-modal interactions and spatial auditory processing
Vision influences how we perceive space by hearing. Ventriloquism effect and after-effect are phenomena illustrating short-term plasticity in spatial hearing induced by visual signals. Visual attentional cuing also influences spatial auditory processing both in terms of sound localization and spatial benefit in speech perception. The current project will examine the effect of visual information on spatial auditory perception by performing behavioral experiments, neuroimaging studies, and computational modeling.
Hladek L, Seitz A, Kopco N (2021) Auditory-visual interactions in egocentric distance perception: Ventriloquism effect and aftereffect. Journal of the Acoustical Society of America, 150, 3593-3607, doi.org/10.1121/10.0007066. Kopčo N, Lokša P, Lin I-F, Groh J, Shinn-Cunningham B (2019). Hemisphere-Specific Properties of the Ventriloquism Aftereffect. Journal of the Acoustical Society of America, 146, EL177 doi.org/10.1121/1.5123176
doc. Ing. Norbert Kopčo, PhD.
Informatics (Id)
Anomaly detection using machine learning methods
Data analysis solutions are applied in various areas of technical, natural, human and economic sciences. Machine learning is a sub-area of artificial intelligence that deals with machine learning methods and algorithms based on input data in a defined solution space. The aim of the dissertation thesis is to design and application of algorithms and methods of machine learning in case studies of anomaly detection and to compare the performance of the proposed solution with other available studies.
The aim of the dissertation thesis is to design and application of algorithms and methods of machine learning in case studies of anomaly detection and to compare the performance of the proposed solution with other available studies.
1. Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM computing surveys (CSUR), 41(3), 1-58. 2. Pang, G., Shen, C., Cao, L., & Hengel, A. V. D. (2021). Deep learning for anomaly detection: A review. ACM computing surveys (CSUR), 54(2), 1-38. 3. Russell, S., & Norvig, P. (2021). Artificial Intelligence: a modern approach, 4th US ed. University of California, Berkeley.
doc. RNDr. Ľubomír Antoni, PhD.
prof. RNDr. Gabriel Semanišin, PhD.
Informatics (IdAj)
Explainable artificial intelligence
In recent years, we have experienced an unprecedented development of artificial intelligence methods and their intensive use in various areas of human society. Along with the use of artificial intelligence, questions arise about the reliability and trustworthiness of these methods and their potential misuse or inadequate use. These issues are particularly sensitive in areas where high reliability is required or where they have a direct impact on human health or social integrity. Indeed, in general, most AI methods are not fully transparent in the way in which the computational and decision-making process is carried out. It is implemented using a large number of parameters, the interpretation of which is very complicated. Explainable Artificial Intelligence (XAI), which aims to develop models and techniques to ensure the explainability of individual algorithms and methods, seeks to eliminate these problems.
The aim of this dissertation is to analyse known approaches and try to develop new explainable artificial intelligence systems that can be considered not only efficient but also trustworthy.
1. Russell, S., & Norvig, P. (2021). Artificial Intelligence: a modern approach, 4th US ed. University of California, Berkeley. 2. Angelov, P. P., Soares, E. A., Jiang, R., Arnold, N. I., & Atkinson, P. M. (2021). Explainable artificial intelligence: an analytical review. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 11(5), e1424. 3. Vilone, G., & Longo, L. (2021). Notions of explainability and evaluation approaches for explainable artificial intelligence. Information Fusion, 76, 89-106.
prof. RNDr. Gabriel Semanišin, PhD.
doc. RNDr. Ľubomír Antoni, PhD.
Informatics (Id)
Forensic analysis of the internet of things
The Internet of Things (IoT) is becoming an integral part of everyday life. Also, it brings a significant increase in security threats and security incidents. An important aspect of the investigation of computer security incidents is an adequate forensic investigation. Within this investigation, several problems can be identified that are related to the heterogeneity of the available IoT-producing components. The aim of the work is to analyze the possibilities of using machine learning methods in securing, extraction and analysis of digital tracks from these devices as well as to design an automated method of extraction and analysis of forensic artefacts from IoT components.
(1) Analyze the possibilities of using machine learning methods in securing and extracting digital traces from IoT devices. (2) Analyze the possibilities of using machine learning methods to analyze digital traces from IoT devices. (3) Design an automated way to extract and analyze forensic artifacts from IoT components.
(1) Stoyanova, Maria, et al. "A survey on the internet of things (IoT) forensics: challenges, approaches, and open issues." IEEE Communications Surveys & Tutorials 22.2 (2020): 1191-1221. (2) Yaqoob, Ibrar, et al. "Internet of things forensics: Recent advances, taxonomy, requirements, and open challenges." Future Generation Computer Systems 92 (2019): 265-275. (3) Atlam, Hany F., et al. "Internet of things forensics: A review." Internet of Things 11 (2020): 100220. (4) Khanpara, Pimal, et al. "Toward the internet of things forensics: A data analytics perspective." Security and Privacy: e306 (2023).
doc. RNDr. JUDr. Pavol Sokol, PhD.
Informatics (Id)
Formal concept analysis
Formal concept analysis is a data-mining method applied to a rectangular matrix of data in which each row corresponds to some object, each column corresponds to some possible attribute, and the matrix field value denotes a membership of the column attribute for row object. One of the goals of this method is to find so-called concepts, which are stable (in some sense) pairs of subsets of objects and attributes. The method can be considered a nice application of the algebraic notion of a Galois connection. It has been described in detail by Ganter and Wille, in particular for the so-called crisp case with binary matrix data. A natural question that arises is what happens if the matrix data are non-binary...
Formálna konceptová analýza je data-miningová metóda na obdĺžnikovej tabuľke, ktorej každý riadok zodpovedá nejakému objektu, každý stĺpec nejakému jeho potenciálnemu atribútu a každé políčko obsahuje informáciu o tom, či (prípadne v akej miere) má príslušný objekty príslušný atribút. Jeden z cieľov tejto metódy je nájsť takzvané koncepty, čo sú v istom zmysle stabilné dvojice podmnožín objektov a atribútov. FCA možno považovať za peknú aplikáciu algebraického pojmu Galoisovej konexie. Pôvodná verzia vychádza z klasického diela Gantera a Willeho a popisuje prípad binárnych dát. Vzniká však prirodzená otázka, čo sa stane, ak údaje v tabuľke nebudú binárne...
prof. RNDr. Stanislav Krajči, PhD.
Informatics (Id)
Intercontext structures and information preserving
Formal Concept Analysis (FCA) provides tools for extracting implicit knowledge from any tabular data. The challenge of this dissertation should be research in the environment of interconnecting multiple tables while maintaining the internal structural and knowledge properties of the input data. The FCA has had theoretical tools for this since its inception, but it does not take into account the semantic side of this problem. But the latest results coming from our institute pave the way which one can find out. It's still just the beginning. The output of this dissertation should be a continuation, whether from a theoretical or algorithmic point of view, they also needed a considerable amount of experimentation with various real data.
Investigate the issue using more abstract theoretical tools such as Category Theory. Support the issue with experimental observations. Testing and interpretation of results on real data.
Michael Barr and Charles Wells. 1990. Category theory for computing science. Prentice-Hall, Inc., USA Bernhard Ganter, Gerd Stumme, and Rudolf Wille (Eds.). 2005. Formal Concept Analysis: foundations and applications. Springer-Verlag, Berlin, Heidelberg.
doc. RNDr. Ondrej Krídlo, PhD.
Informatics (Id)
Intuitionistic fuzzy sets and their applications
Intuitionistic fuzzy sets (IFS) have emerged as a powerful extension of classical fuzzy sets, offering a more flexible framework for dealing with uncertainty and vagueness in real-world problems. This thesis aims to provide a comprehensive exploration of IFS properties and their wide-ranging applications. It will delve into the theoretical aspects of IFS, examine various applications across different domains, present case studies to illustrate practical implementation, evaluate performance through experimentation, and discuss future directions and challenges in the field.
This thesis aims to provide a comprehensive exploration of IFS properties and their wide-ranging applications. It will delve into the theoretical aspects of IFS, examine various applications across different domains, present case studies to illustrate practical implementation, evaluate performance through experimentation, and discuss future directions and challenges in the field.
1. Atanassov, K. T. (2012). On intuitionistic fuzzy sets theory (Vol. 283). Springer. 2. Atanassov, K. T., & Atanassov, K. T. (1999). Intuitionistic fuzzy sets (pp. 1-137). Physica-Verlag HD. 3. Kridlo, O., & Ojeda-Aciego, M. (2017, July). Extending formal concept analysis using intuitionistic L-fuzzy sets. In 2017 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) (pp. 1-6). IEEE.
doc. RNDr. Ondrej Krídlo, PhD.
doc. RNDr. Ľubomír Antoni, PhD.
Informatics (IdAj)
Intuitionistic fuzzy sets and their applications
Intuitionistic fuzzy sets (IFS) have emerged as a powerful extension of classical fuzzy sets, offering a more flexible framework for dealing with uncertainty and vagueness in real-world problems. This thesis aims to provide a comprehensive exploration of IFS properties and their wide-ranging applications. It will delve into the theoretical aspects of IFS, examine various applications across different domains, present case studies to illustrate practical implementation, evaluate performance through experimentation, and discuss future directions and challenges in the field.
This thesis aims to provide a comprehensive exploration of IFS properties and their wide-ranging applications. It will delve into the theoretical aspects of IFS, examine various applications across different domains, present case studies to illustrate practical implementation, evaluate performance through experimentation, and discuss future directions and challenges in the field.
1. Atanassov, K. T. (2012). On intuitionistic fuzzy sets theory (Vol. 283). Springer. 2. Atanassov, K. T., & Atanassov, K. T. (1999). Intuitionistic fuzzy sets (pp. 1-137). Physica-Verlag HD. 3. Kridlo, O., & Ojeda-Aciego, M. (2017, July). Extending formal concept analysis using intuitionistic L-fuzzy sets. In 2017 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) (pp. 1-6). IEEE.
doc. RNDr. Ondrej Krídlo, PhD.
doc. RNDr. Ľubomír Antoni, PhD.
Informatics (Id)
Cross-modal interactions and spatial auditory processing
Vision influences how we perceive space by hearing. Ventriloquism effect and after-effect are phenomena illustrating short-term plasticity in spatial hearing induced by visual signals. Visual attentional cuing also influences spatial auditory processing both in terms of sound localization and spatial benefit in speech perception. The current project will examine the effect of visual information on spatial auditory perception by performing behavioral experiments, neuroimaging studies, and computational modeling.
Hladek L, Seitz A, Kopco N (2021) Auditory-visual interactions in egocentric distance perception: Ventriloquism effect and aftereffect. Journal of the Acoustical Society of America, 150, 3593-3607, doi.org/10.1121/10.0007066. Kopčo N, Lokša P, Lin I-F, Groh J, Shinn-Cunningham B (2019). Hemisphere-Specific Properties of the Ventriloquism Aftereffect. Journal of the Acoustical Society of America, 146, EL177 doi.org/10.1121/1.5123176
doc. Ing. Norbert Kopčo, PhD.
Informatics (IdAj)
Modeling and algorithms for construction of smooth curves
Recently we proposed a new approach to solving the tridiagonal systems on a uniform grid of nodes. One of the goals of the thesis is to investigate the effect of the given approach on the nonuniform grid and the inversion of tridiagonal matrices. Classic cubic splines of class C2 are implicit. We succeeded to express them in an explicit form that enables a design of a linear model for approximation and estimate of spline coefficients. The second goal is to analyze the properties of LS estimate of coefficients and their comparison with B-splines. The third goal is finding of criterion of optimal stopping in on-line approximation according to prediction in an appropriate metric.
The goal is proposing models and algorithms for global and sequential parametric approximation, interpolation, smoothing and automatic node detection.
Literatúra - Salomon, D.: Curves and Surfaces for Computer Graphics, Springer, 2006. - Kačala V., Török Cs., Speedup of tridiagonal system solvers, Journal of Computational and Applied Mathematics 381 (2021) 112997 - J.Hudák, Cs.Török, Ľ.Antony, Explicit forms of interpolating cubic splines and data smoothing, to appear
doc. RNDr. Csaba Török, CSc.
Informatics (Id)
Modeling and algorithms for construction of smooth curves
Recently we proposed a new approach to solving the tridiagonal systems on a uniform grid of nodes. One of the goals of the thesis is to investigate the effect of the given approach on the nonuniform grid and the inversion of tridiagonal matrices. Classic cubic splines of class C2 are implicit. We succeeded to express them in an explicit form that enables a design of a linear model for approximation and estimate of spline coefficients. The second goal is to analyze the properties of LS estimate of coefficients and their comparison with B-splines. The third goal is finding of criterion of optimal stopping in on-line approximation according to prediction in an appropriate metric.
The goal is proposing models and algorithms for global and sequential parametric approximation, interpolation, smoothing and automatic node detection.
Literatúra - Salomon, D.: Curves and Surfaces for Computer Graphics, Springer, 2006. - Kačala V., Török Cs., Speedup of tridiagonal system solvers, Journal of Computational and Applied Mathematics 381 (2021) 112997 - J.Hudák, Cs.Török, Ľ.Antony, Explicit forms of interpolating cubic splines and data smoothing, to appear
doc. RNDr. Csaba Török, CSc.
Informatics (Id)
Plasticity and attention in spatial hearing
In everyday situations, humans are exposed to multiple concurrent stimuli in complex, continuously changing environments. To correctly extract relevant information, they adapt their processing to reflect the specifics of the current scene, and they learn from previous experience to improve the perceptual strategies used. The current project proposes to perform a series of behavioral experiments, brain imaging studies, and computational modeling to study how attention and mechanisms of implicit and explicit learning are used to cope with complex listening environments for speech processing, sound localization, and learning of new phonetic categories.
Vlahou E, Ueno K, Shinn-Cunningham B, Kopco N (2021) Calibration of consonant perception to room reverberation. Journal of Speech, Language, and Hearing Research, 64(8), 2956-2976 Klingel M, Laback B, Kopco N (2021) Reweighting of Binaural Localization Cues Induced by Lateralization Training. Journal of the Association for Research in Otolaryngology, 22, 551–566, https://doi.org/10.1007/s10162-021-00800-8.
doc. Ing. Norbert Kopčo, PhD.
Informatics (IdAj)
Plasticity and attention in spatial hearing
In everyday situations, humans are exposed to multiple concurrent stimuli in complex, continuously changing environments. To correctly extract relevant information, they adapt their processing to reflect the specifics of the current scene, and they learn from previous experience to improve the perceptual strategies used. The current project proposes to perform a series of behavioral experiments, brain imaging studies, and computational modeling to study how attention and mechanisms of implicit and explicit learning are used to cope with complex listening environments for speech processing, sound localization, and learning of new phonetic categories.
Vlahou E, Ueno K, Shinn-Cunningham B, Kopco N (2021) Calibration of consonant perception to room reverberation. Journal of Speech, Language, and Hearing Research, 64(8), 2956-2976 Klingel M, Laback B, Kopco N (2021) Reweighting of Binaural Localization Cues Induced by Lateralization Training. Journal of the Association for Research in Otolaryngology, 22, 551–566, https://doi.org/10.1007/s10162-021-00800-8.
doc. Ing. Norbert Kopčo, PhD.
Informatics (Id); Informatics (Ide)
Problems of implementing cryptographic primitives in network protocols.
Today's secure Internet network protocols simultaneously use several cryptographic primitives in different combinations and in different modes. Although the individual primitives have been subjected to many security analyses, their simultaneous use can lead to additional unexpected vulnerabilities. This research work will analyze the possibilities of parallel attacks on primitives, the cryptographic functions used and their modes of use with the intention of proving the achievability of specific security goals, and possibly proposing protection against the vulnerabilities identified.
Analysis of the possibility of parallel attacks on cryptographic primitives used simultaneously in multiple network protocols executing in parallel. Proofs of achievement of security objectives, design of defenses against identified vulnerabilities.
1. P.Y.A. Ryan, S.A.Schneider: Modelling and Analysis of Security Protocols, Addison Wesley, 2001 2. C. Boyd, A. Mathuria: Protocols for Authentication and Key Establishment, Springer, 2003 3. G. Bella: Formal Correctness of Security Protocols, Springer 2007.
doc. RNDr. Jozef Jirásek, PhD.
Informatics (Id)
Shannon's sampling theorem and real problems
The sinc function based sampling theorem provides sufficient conditions to guarantee that infinite discrete sequences capture all the information from the continuous signal and thus enables the full reconstruction of the original one. Thanks to it, it is possible to develop various geometric, physical or numerical models for practice. However, observed signals, sequences of real records or simulated and calculated data are final, and practice has shown that for their effective description, interpolation, approximation or analysis it is insufficient to consider only sinc functions, especially in the case of uniform nodes. The question is how to design the models so that they can be used effectively in applications and give the most accurate results.
The goal is to improve Shannon's sampling theorem by extended model change, optimal selection of uniform nodes and damping parameters.
- M.Kircheis, D.Potts, Manfred Tasche, On regularized Shannon sampling formulas with localized sampling, Sampling Theory, Signal Processing, and Data Analysis, 2022, Springer - M.Richardson, L.N. Trefethen, A SINC FUNCTION ANALOGUE OF CHEBFUN, SIAM J. SCI. COMPUT., 2011, Vol. 33, No. 5, pp. 2519–2535 - R.Platte, L.N.Trefethen, A.B.J.Kuijlaars, Impossibility of fast stable approximation of analytic functions from equispaced samples, SIAM REVIEW, Society for Industrial and Applied Mathematics, 2011, Vol.53, No.2, pp.308–318
doc. RNDr. Csaba Török, CSc.
Informatics (IdAj)
Shannon's sampling theorem and real problems
The sinc function based sampling theorem provides sufficient conditions to guarantee that infinite discrete sequences capture all the information from the continuous signal and thus enables the full reconstruction of the original one. Thanks to it, it is possible to develop various geometric, physical or numerical models for practice. However, observed signals, sequences of real records or simulated and calculated data are final, and practice has shown that for their effective description, interpolation, approximation or analysis it is insufficient to consider only sinc functions, especially in the case of uniform nodes. The question is how to design the models so that they can be used effectively in applications and give the most accurate results.
The goal is to improve Shannon's sampling theorem by extended model change, optimal selection of uniform nodes and damping parameters.
- M.Kircheis, D.Potts, Manfred Tasche, On regularized Shannon sampling formulas with localized sampling, Sampling Theory, Signal Processing, and Data Analysis, 2022, Springer - M.Richardson, L.N. Trefethen, A SINC FUNCTION ANALOGUE OF CHEBFUN, SIAM J. SCI. COMPUT., 2011, Vol. 33, No. 5, pp. 2519–2535 - R.Platte, L.N.Trefethen, A.B.J.Kuijlaars, Impossibility of fast stable approximation of analytic functions from equispaced samples, SIAM REVIEW, Society for Industrial and Applied Mathematics, 2011, Vol.53, No.2, pp.308–318
doc. RNDr. Csaba Török, CSc.
Informatics (Id)
Attacks on machine learning methods in the field of cybersecurity
Machine learning methods play an essential role in responding to security incidents. To detect security incidents, respectively security attacks, these methods make training data models of normal behaviour and detect incidents, respectively attacks as deviations from these models. This process encourages the attackers to manipulate training data in such a way that the learned model cannot detect their subsequent attacks. In addition to the learning phase, security systems using machine learning methods are also vulnerable to various attacks during the decision-making phase. Using specially selected inputs, the attacker bypasses the learned behaviour of the detection system. This work aims to analyze used machine learning methods in the field of cybersecurity concerning their resistance to the above attacks. At the same time, the aim is to propose a method of testing machine learning methods with regard to the possibility of their misuse by the attacker and how to protect these methods against various types of attacks.
(1) Analyze the used machine learning methods in cyber security concerning their resistance to attacks. (2) To propose a method of testing machine learning methods considering the possibility of their abuse by an attacker. (3) Design a way to protect machine learning methods against attacks.
(1) Debicha, Islam, et al. "Adv-Bot: Realistic Adversarial Botnet Attacks against Network Intrusion Detection Systems." Computers & Security (2023): 103176. (2) Pawlicki, Marek, Michał Choraś, and Rafał Kozik. "Defending network intrusion detection systems against adversarial evasion attacks." Future Generation Computer Systems 110 (2020): 148-154. (3) Richards, Luke E., Edward Raff, and Cynthia Matuszek. "Measuring Equality in Machine Learning Security Defenses." arXiv preprint arXiv:2302.08973 (2023).
doc. RNDr. JUDr. Pavol Sokol, PhD.
Informatics (Id)
Explainable artificial intelligence
In recent years, we have experienced an unprecedented development of artificial intelligence methods and their intensive use in various areas of human society. Along with the use of artificial intelligence, questions arise about the reliability and trustworthiness of these methods and their potential misuse or inadequate use. These issues are particularly sensitive in areas where high reliability is required or where they have a direct impact on human health or social integrity. Indeed, in general, most AI methods are not fully transparent in the way in which the computational and decision-making process is carried out. It is implemented using a large number of parameters, the interpretation of which is very complicated. Explainable Artificial Intelligence (XAI), which aims to develop models and techniques to ensure the explainability of individual algorithms and methods, seeks to eliminate these problems.
The aim of this dissertation is to analyse known approaches and try to develop new explainable artificial intelligence systems that can be considered not only efficient but also trustworthy.
1. Russell, S., & Norvig, P. (2021). Artificial Intelligence: a modern approach, 4th US ed. University of California, Berkeley. 2. Angelov, P. P., Soares, E. A., Jiang, R., Arnold, N. I., & Atkinson, P. M. (2021). Explainable artificial intelligence: an analytical review. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 11(5), e1424. 3. Vilone, G., & Longo, L. (2021). Notions of explainability and evaluation approaches for explainable artificial intelligence. Information Fusion, 76, 89-106.
prof. RNDr. Gabriel Semanišin, PhD.
doc. RNDr. Ľubomír Antoni, PhD.