Information and communication technology

1. The use of innovative machine learning methods, including associative and motivated learning, to create intelligent, cognitive, self-learning agents.

Supervisor: dr. hab. Horzyk Adrian, prof. AGH
Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract:  The research issue includes the development of new and extension of existing models of broadly understood artificial intelligence, computational intelligence, knowledge engineering in order to enable automatic adaptation, creation of knowledge and intelligent behavior of agents. Research work will also compare the developed methods, models and algorithms to those currently used. The comparative criterion will be the agents' ability to self-organize and create a knowledge base based on their interaction with the environment. After the classic models and methods of machine learning, the implementation of the problem will be applied to associative data structures and their neural networks and motivated learning methods.

Research facilities: The research unit and the supervisor have knowledge and experience in the creation and development of methods of artificial intelligence, computational intelligence and knowledge engineering, so they have the ability to carry out research within scientific projects, including the above mentioned research topic.

Number of places: 2

 

2. Applying combinatorial optimization methods to modelling and implementation of parallel systems.

Supervisor: dr. hab. inż. Andrei Karatkevich, prof. AGH
Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract:A parallel system can be represented as a composition of the sequential systems. For example, a hierarchical FSM can be seen as a network of communicating classical FSMs, and a Petri net – as a composition of the State Machine nets. Such decomposition can help in conversion between different kinds of models, simplify formal analysis, and it may be necessary for the system implementation (if, in general, a parallel system has to be implemented as a set of communicating sequential devices, actors or processes). However, the tasks of such decomposition are connected with a series of problems, partially unsolved. One of them is possibility of the decomposition and its conditions. For example, not every Petri net can be covered by the State Machine subnets, and if it can be covered, then sometimes, introducion of the additional places is required. Conditions of existence of such cover are not known well enough.

Another related topic is minimization of the number of sequential components covering a parallel system. In some cases and for some models such decompositional is evident, but in other cases – for the Petri nets, for example – multiple variants with different number of components may exist, and the problem of minimization arises. It is similar to the set cover problem, which is NP-hard in general case, hence it makes sense to design and use the corresponding approximation algorithms. On the other hand, there are cases in which the minimum cover can be found in polynomial time. So, research on the conditions of such situation and appropriate algorithms is essential.

Summarizing, the interesting research problems are connected with conditions of sequential decomposition of the concurreny systems, algorithms of such decomposition, and conditions of possibility of finding the minimum cover quickly (in polynomial time).

Research facilities: Department of Applied Computer Science of AGH has hardware and software resources necessary for research in the mentioned area. The Department and the supervisor have got experience in formal modelling and analysis of concurrent systems.

Number of places: 2

 

3. Methods for intelligent analysis of data streams for explainability.

Supervisor: prof. dr hab. inż. Grzegorz J. Nalepa

Auxiliary supervisor: dr inż. Szymon Bobek, dr inż. Krzysztof Kutt
Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract:  Currently Artificial Intelligence (AI) methods are commonly used to analyze large data sets. The sources of this data depend on the domain and may include sensors in inustrial installations, wearable sensors, and ambient intelligence systems. The use of AI methods allows for solving new problems such as sdaptaion of systems, personalization of services, or predictive maintanance. Today data acqusition is often not an important challange anymore. We also have a range of state-of-the art methods for prediction and classification based on the data. However, a growing and very important challange is in fact related to the proper explanation of the operation of the models produced by these methods. Furthemore, we want to be able to interpret and assess the impact of the changes in the data on the operation of these models, and finally the data bias as well. The research in this task will address data from diverse domains including Industry 4.0, video games, or physiological data from wearable sensors for affective computing applications. We expect from the PhD candidate basic knowledge of AI methods, high motivation and engagement.

Research facilities: The GEIST.re research team operating in the Institute of Applied Computer Science at the EAIiIB faculty has a long experience in projects in both fundamental (e.g. NCN) as well as applied (e.g. NCBR, MCP) research. The team members have co-authored many research papers in top international journals. Currently the team coordinates an international project „CHIST-ERA PACMEL: Process-aware Analytics Support based on Conceptual Models for Event Logs” related to this research task. The team can also provide appropriate office space for PhD researchers.

Number of places: 2

 

 

4. Determination of the vehicle velocity simulation model based on measurements of selected route parameters

Supervisor: dr hab. inż. Adam Piórkowski

Auxiliary supervisor: dr inż. Edyta Kucharska

Department of Biocybernetics and Biomedical Engineering, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The aim of the research is to develop the vehicle speed model based on selected route parameters and historical data. The scientific problems that are provided here are the development of new algorithms to reduce the time of the computational process in relation to known solutions while maintaining an acceptable result. It is planned to design an appropriate computing environment and develop methods to optimize time of data processing, including parallelization of selected tasks.

 

Research facilities: Historical GPS data are available for the research needs

Number of places: 1

 

5. Graphical Modeling: BPMN, UML, DMN, DG, SSG, PN, BN. Methods, Tools, and Applications

Supervisor: prof. dr hab. inż. Antoni Ligęza

Auxiliary supervisors: dr inż. Krzysztof Kluza, dr Krystian Jobczyk, dr inż. Weronika T. Adrian,

Department of Applied Computer Science, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Abstract: The research concerns theory, tools and applications of graphical modeling with: BPMN, DMN, UML, Decision Graphs, State-Space Graphs, Petri Nets,Bayesian Networks, etc. The research is planned to be focused on: (I) analysis of graphical models and their properties, verification and validation of such models, , automated synthesis and optimization, and mutual proliferation/synergy effects. It is planned to consider development of applications and their applications (in Business Processes, Intelligent Control, Data Analysis, Decision Making, etc.).

Research facilities: The Department has access to all resources offered by the Central Library of AGH, as well as to internet. We also have intellectual capital, creative potential and experience in research at the international level

Number of places: 3

 

6. Constraint Programming, Logic and Probabilistic Programming. Methods, Tools, and Applications

Supervisor: prof. dr hab. inż. Antoni Ligęza

Auxiliary supervisors: dr inż. Krzysztof Kluza, dr Krystian Jobczyk, dr inż. Weronika T. Adrian,

Department of Applied Computer Science, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Abstract: It is planned to focus the research on theory, tools, and applications of Constraint Programming and Logic and Probabilistic Programming. The most interesting areas of research include: (i) improvement of reasoning efficiency with use of granularization, constructive abduction, smart decomposition, machine learning and heuristics, and the notion of entropy and causal graphs, (ii) appliactions among other in structure (graph) analysis and synthesis, planning and scheduling, intelligent control, and (iii) development of tools based on existing systems (such as MiniZinc, Prolog, ASP, Python and its libraries, Problog, Bayesian Networks, etc.).

Research facilities: The Department has access to all resources offered by the Central Library of AGH, as well as to internet. We also have intellectual capital, creative potential and experience in research at the international level

Number of places: 3

 

 

7. New methods for developing security architectures for complex Information systems

Supervisor: Prof. dr hab. Marek R. Ogiela

Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Abstract: The research problem covers design of new and development of existing models of creating and maintain security architectures for complex Information Systems – both on the level of specific IT Solutions (such as singular IT System or Internet protocols) and enterprise architecture level. The special focus will be laid on the systems implementing and/or utilizing emerging technologies such as: artificial intelligence, distributed ledger (e.g. blockchain), cloud, or Internet-of-Things. The research work will concentrate around designing and analysing new methods for developing the security architecture for such systems but also comparing those to the existing industry standards (e.g. COBIT, SABSA, TOGAF, NIST). The analysis will be also taken around emerging security threats related to the usage of these new technologies (e.g. computer malware utilizing the Artificial Intelligence algorithms), as well as effective methods for protecting against those, that are defined on the level of system’s architecture.

Research facilities: The Scientific Unit and the supervisor have the necessary knowledge and experience in designing and developing complex Information Systems and Technologies (such as Artificial Intelligence systems), as well as designing and embedding security in such systems (e.g. cryptographic systems). The unit is capable of running the research work incl. the aforementioned research problem.

Number of places: 1

 

8. Analysis and classification of metal and alloy microstructures using machine learning tools

Supervisor: Prof. dr hab. inż. Jan Kusiak

Department of Applied Computer Science and Modelling, Faculty of Metals Engineering and Industrial Computer Science

Abstract: The main task of forming is directed change of the shape and mechanical and strength properties of deformed metals. Therefore, one of the most important tasks of metal forming is to coordinate the transition between the forming step and heat treatment in a way that ensures that the formed semi-finished product or finished product is created with minimal energy and at the same time meets the customer's quality requirements. The mechanical and strength properties of metals and alloys have a major impact on their microstructure, which in turn results from the technological parameters of forming (temperature, deformation, deformation speed, etc.) and changes occurring during thermo-mechanical treatment. Therefore, in order to precisely control the properties of metal, constant and precise supervision over its microstructure in the whole production cycle is necessary. Existing microstructure measurement methods are mainly focused on simple algorithms that are based on the subjective classification of a human being, which may be fraught with an unsystematic error. Therefore, the main research task is to apply techniques taken from the dynamically developing field of image analysis based on machine learning, and in particular on Convolutional Neural Networks, for the analysis and classification of metal microstructures. The developed algorithms may not only have a theoretical value, but also a practical one. They can be used in systems for automatic evaluation and classification of microstructures, necessary in the control of metal forming technologies.

Research facilities: The research will mainly base on the computer infrastructure existing in the Department of Applied Computer Science and Modeling and the obtained microstructures data from institutions cooperating with the Department (IMŻ in Gliwice, BA Freiberg, etc.). The preliminary results obtained in this field confirm the great potential of this subject.

Number of places: 1

 

9. Approximative modelling and verification of software

Supervisor: Prof. dr hab. inż. Tomasz Szmuc

Department of Applied Computer Science , Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Rapid development of embedded systems applications, an increase of their functionality in the framework of energy restrictions and high level of concurrency – all these features make reasonable searching for new modelling and verification methods supporting software development. The main goal of the research is a development of modelling and verification methods to support the development of correct software. Markow processes and probabilistic logics will be used in the initial state of the research leading to quantitative (in opposition to currently used qualitative) proving of the correctness. For example, quantitative means that system is correct in the range of 95%. The two main research topics are planned:

  • Modelling and analysis of logical correctness
  • Verification of time constraints satisfiability.

Research facilities: Research in the area of formal modelling and correctness analysis have reached, in the Department of Applied Computer Science, an exceptional level resulting in essential achievements. Several models based on formal description, i.e. Petri nets, process algebras (LOTOS) and temporal logics have been developed. The additional area related to translations from engineering development thread (UML, SysML, AADL) into above mentioned formal models should be indicated. The proposed area is then a natural continuation of the research.

Number of places: 2

 

10. Detecting Clusters by Adaptive Algorithms in The Context of Cloud Security Profiles.

Supervisor: dr hab. inż. Lucjan Janowski

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Detecting Clusters by Adaptive Algorithms in The Context of Cloud Security Profiles is for anomaly-based detection system running on the endpoints (client computers, servers) which is building profiles for process, files and network behavior. Those profiles can be stored in a Cloud and used as a Cloud service. A deviation from the profile generates the anomaly which should be investigated. Currently, most of the solutions we have used to protect endpoints based on well-known signatures which are stored locally on the endpoint or in the Cloud, for example, IPS rules of well-known threats, hashes of well-known malicious files, reputation of IP addresses for incoming and outgoing traffic connections, or domain and URL reputation. These signatures have no ability to detect new attacks, zero days attacks, and even some attacks like attacks targeting Web Applications or Database. We have in Cisco a solution which is called “Cisco Cognitive Threat Analytics– CTA” which is network based. CTA is using cloud service to group similar patterns (netflows, syslogs from web proxy) in clusters. Those which land in a cluster which resembles malicious behavior generates anomaly. We still don’t have such solution for endpoints. The innovation of this idea is to create an algorithm for a clustering profile for every server and application (via profiling – gathering statistical data) and classify those into the clusters in the Cloud by using metadata (statistics per process, files, and network behavior) to build profile’s clusters and correlate collected data with other users to determine anomaly behavior. Once a certain server is compromised it could start presenting the deviation from the profile and generate anomaly. The goal is to use machine learning to utilize Cloud service and Big Data to build smarter profiles. This proposal is for anomaly-based detection system running on the endpoints (client computers, and servers) which is building profiles for process, files and network behavior. Those profiles can be stored in a Cloud and used as a Cloud service. A deviation from the profile generates the anomaly which should be investigated. The Clustering profiles can be classified by computing a new representation that leverages all flows in a profile to capture malware dynamics and behavior in time. The representation is robust to malware variations attempting to evade detection (e.g. by changing the URL pattern, number of transferred bytes, user agent, etc.). The invariant representation is based on the idea that malicious flows in a profile will have different statistical properties than legitimate flows in another profile, the representation is used in a clustering to group malware belonging to the same category. The conceptual problem is removing noisy variables from large number of possible node indicators. In unsupervised machine learning, such as clustering, such noisy data cancan easily remove relation between different clusters. Depending on the attack profile different variables can be noisy or meaningful. Our analysis will focus on detecting the important variables. We will also use supervised training approach where we can mark good and bad clusters or even certain behaviors to help the system learning and tuning the classification process. Clustering has been used to detect intrusion, in IaaS (Infrastrure as a Service) cloud monitoring, and to improve kernel clustering algorithm for intrusion detection. My research will focus on crating an algorithm for clustering profiles in such way that anomaly detection is possible.

Research facilities: Cisco is the research facility and we have available data from AMP for Endpoints cloud, laboratories to perform some attacks, and all needed resources to make this research successful.

Number of places: 1

 

11. Adaptive Multihoming in IP networks.

Supervisor: dr hab. inż. Marcin Niemiec

Auxiliary supervisor:

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The topic of research are adaptive (i.e. adjustable to a current network load) algorithms of multihoming interface selection. The specific goal is development of an algorithm of outgoing interface selection for subsequent network flows, which will maximize achieved throughput or minimize delay. Parameters of existing network flows will be sampled continuously. Basing on these observations, quality of paths to specific destination networks (AS) on specific interfaces will be estimated. When a new flow will appear, interface which currently yields a better path (according to a given policy) to a given destination network will be chosen. Implementation aspect of this PhD research topic will be implementation of the developed algorithm in home router (based on Linux kernel), which will be launched onto market as a new product.

Research facilities: Department of Telecommunications owns hardware required for research on the topic. PhD student will have an access to wired network lab and wireless computer networks lab. Department owns several multiprocessing servers, which can be used for simulations and emulation of virtual networks with implemented algorithms. Apart from that, PhD student will have a possibility to use PL-GRID computing infrastructure.

Number of places: 1

 

12. A system of automated purchase of advertising space in real time in the auction model.

Supervisor: dr hab. Konrad Kułakowski

Auxiliary supervisor:

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: In the existing advertising sales systems customers of an internet portal – advertisers – provide a set of advertisements for presenting to the users of the portal. However, they cannot be sure which of the prepared advertisements will be presented to the visitors of the portal. This is because some advertisements may never match the profiles of the users watching the given content. Such a situation causes the loss for the advertisers, who have to create more advertising content that they need. The RTB (the real-time bidding) model may come to the rescue. According to this approach, the internet portal provides to the advertisers an automatized auction platform allowing them to compete for advertising space connected with the given single advertisement. Adopting this model enables the advertiser to reduce the costs spent on creation, not showing ads, while the internet portal can optimize the advertising space prices. Implementation of such a solution, however, requires providing the advertiser with infrastructure that allows him to create his software agent taking part in the auction and making decisions based on the available data. In particular, the agent needs to have the contextual knowledge about the profiles of the users which may be interested in the given advertisement. Thus, the internet portal has to provide the advertiser with the ready-to-use data analysis algorithms, data, and the mechanisms allowing the agent is making decisions during the auctions. The implementation of such a system will increase the competitiveness of the internet portal as a provider of advertising space and give advertisers a flexible tool of cost reduction. Thanks to the proposed solution, the number of unwanted and irrelevant advertisements will be reduced, thus increasing the comfort of using our internet portal.

Research facilities: In order to implement the project from the university, a PhD candidate needs periodic access to a computer with the Linux operating system and access to the Internet. Access to the web portal system is provided by the company.

Number of places: 1

 

13. Distributed system for exchange of different cryptocurrency blocks between users .

Supervisor: prof. dr hab. inż. Tomasz Szmuc

Second supervisor: dr hab. inż. Igor Wojnicki

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The proposed research is focused on the development of blockchain technology by transition from centralized servers centralized into distributed systems servicing crypto-currency accounts. Existing solutions require a third party for controlling exchange between the related users. This aspect leads to several drawbacks, i.e. increasing of exchange time, disclosure of anonymity, etc. Lastly proposed new solutions (Atomic Swap primitives) enable flexible exchange of cryptocurrency without interconnection via a centralized stock exchange, third parties, connecting tokens. The exchange is carried out via payment channels, and every transaction needs the only activation of channel enabling payments in both currencies. It seems that the solution is up-and-coming, but currently reached the early development stage, far from commercial applications. The goal of the research is an analysis of Atomic Swap primitive and its implementation enabling exchange of different cryptocurrency blocks in the way independent on cryptographic algorithms, and preserving several features of blockchain technology, i.a. safety, complete anonymity of transactions, etc. Implementation of protocols and algorithms will be cross-related with formal modelling and proving of the required properties.

Research facilities: Research in the area of software engineering, development of specialized software, and formal modelling and correctness analysis have reached, in the Department of Applied Computer Science, an exceptional level resulting in essential achievements. Several models based on formal description, i.e. Petri nets, process algebras (LO-TOS) and temporal logics have been developed, and the models were used in the formal analysis of software. The proposed area is then a natural continuation of the research.

Number of places: 1

 

14. Agent-based computational metaheuristics in solving hard problems.

Supervisor: dr hab. inż. Aleksander Byrski, prof. AGH

Auxiliary supervisor: dr inż. Kamil Piętak

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Agent algorithms are popular methods for solving difficult problems. Metaheuristics, such as PSO or ACO have significant agentistics, in the Department of Computer Science for years, work is underway on the Evolutionary Multi-agent System and also on dedicated computing environments. Such algorithms are also the basis for a number of hybrid solutions. The work carried out in this topic will focus on developing new agent metaheuristic algorithms, also hybrid ones, based on the above-mentioned and similar computational methods.

Research facilities: AGH UST Department of Computer Science has fully equipped laboratories dedicated to research in the field of bio-inspired agent-based artificial intelligence algorithms and workplaces for Ph.D. students. The Academic Computer Centre Cyfronet AGH has the resources necessary to conduct computational and simulation experiments in the field of bio-inspired agent-based artificial intelligence algorithms. Ph.D. students have access to the resources of the Prometheus and Zeus supercomputers.

Number of places: 1

 

15. Intelligent management of traffic in multi-layer Software-Defined Networks.

Supervisor: dr hab. inż. Jerzy Domżał

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The research issue concerns the development of intelligent traffic control methods in multi-layer software-defined networks. In particular, the developed solutions are to provide traffic service with a guaranteed quality of service in multi-layer networks with a central controller. The basic assumption is to ensure cooperation between the optical layer, IP and application layer in order to handle traffic in accordance with the set requirements. In detail, it will be necessary to develop solutions enabling reliable operation of the network, as well as supporting methods of traffic classification and handling of flows with specified quality.

Research facilities: The Department of Telecommunications has appropriate research facilities enabling the conducting of research within scientific projects. The doctoral student will have a place to work, a computer and access to the laboratory. The doctoral dissertation will be carried out as part of the project "Intelligent management of traffic in multi-layer Software-Defined Networks" financed by NCN under the OPUS programme. An additional scholarship of PLN 4,500 was planned for a Ph.D. student for a period of 12 months. There is also the possibility of obtaining other projects ensuring further financing of the scholarship.

Number of places: 1

 

16. Bio-inspired agent-based artificial intelligence algorithms for economic and financial problems.

Supervisor: dr hab. inż. Rafał Dreżewski

Auxiliary supervisor: dr Sylwia Kruk

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The innovative concept of bio-inspired multi-agent systems provides the mechanisms of integrating, and thus more fully exploiting of the potential effect of synergy, several approaches that are currently being intensively developed and are the focus of many researchers: multi-agent computing systems, agent-based approach to modeling and simulation, and bio-inspired artificial intelligence algorithms, both agent and non-agent ones. The results of the conducted research indicate that the bio-inspired multi-agent systems perform very well in the case of multi-modal, non-stationary and multi-objective problems as well as in the tasks of generating investment strategies and multi-objective optimization of the investment portfolio. They are characterized by high generalization capacity and can generate investment strategies and economic and financial decisions that are rational, and at the same time original and non-obvious, which is particularly important in the case of an unstable and rapidly changing socio-economic environment. The main objective of the research will be the development and verification of innovative agent-based bio-inspired artificial intelligence algorithms, intended for solving selected problems in the field of finance and economics, for which optimal solutions have not been found yet. The planned research will concern algorithms for problems such as optimization of the company's capital structure aimed at long-term maximization of its value, multi-objective optimization of the capital structure of the enterprise including sustainable development goals and optimization of the deterministic model of enterprise value growth leading to maximization of the company's value.

Research facilities: AGH UST Department of Computer Science has fully equipped laboratories dedicated to research in the field of bio-inspired agent-based artificial intelligence algorithms and workplaces for Ph.D. students. The Academic Computer Centre Cyfronet AGH has the resources necessary to conduct computational and simulation experiments in the field of bio-inspired agent-based artificial intelligence algorithms. Ph.D. students have access to the resources of the Prometheus and Zeus supercomputers.

Number of places: 1

 

17. AI utilization in virtualized computing and communication infrastructure management and autonomic configuration of native cloud applications supporting 5G systems.

Supervisor: prof. dr hab. inż. Krzysztof Zieliński

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Contemporary computing and communication infrastructure is characterized by a high degree of virtualization and increasing complexity. This also applies to applications that are currently implemented as native cloud applications, i.e. in the form of scaled graphs of communicating functional components. These applications can apply to both the system and business layers. The system layer is connected with creating virtual network functions on demand and communication networks with guaranteed quality (network slicing) connecting core clouds with edge clouds. The system layer constructed in this way is characterized by high flexibility and configurability that allows it to be adapted to the requirements of business applications. Business applications must be appropriately allocated to the system layer elements, so as to finally ensure high requirements in terms of low response times, high throughput, scalability and fault tolerance in dynamically changing load conditions. These requirements are particularly high for 5G systems. The presented technical issue is very complex and must be solved on demand, hence it is believed that its solution must be supported by AI methods. The implementation of research in this area is the subject of the proposed research topic.

Research facilities: There is a possibility to develop the proposed topics within the framework of scientific and research cooperation with Samsung Research Poland and CISCO Systems. These companies have their branches in Krakow. In recent years, several grants financed by Samsung regarding cloud computing organizations have been implemented at the IT Department of WIEiT. Currently, a grant application is being prepared jointly with this company related to the topic of building applications for 5G systems. The CISCO company is also very interested in cooperation with AGH in the area of ​​AI. Together with this company, the "Małopolska Educational Cloud" project is currently being implemented. This is a good starting point for further cooperation with CISCO Systems.

Number of places: 1

 

18. Algorithms for the deployment of sensor data processing on the edge of the network for the Internet of Things systems.

Supervisor: prof. dr hab. inż. Krzysztof Zieliński

Auxiliary supervisor: dr inż. Tomasz Szydło

Faculty of Computer Science, Electronics and Telecommunications

Abstract: One of the issues related to the construction of Internet of Things systems is the processing of sensor data close to their source - on the edge of the Internet network in order to, among others, increase the system's responsiveness, reduce network traffic and preserve energy resources. The aim of the research is the development of simulation and modeling tools for Internet of Things systems operating on the edge of the network, and then development of the algorithms for the deployment of sensor data processing on the available hardware infrastructure to ensure the desired system properties. Research work will include a series of experiments to analyze the operation of typical Internet of Things systems, their modeling and analysis. The developed solutions will be the basis for the tools for a new generation of Internet of Things systems.

Research facilities: The WIET Department provides the necessary infrastructure for conducting research.

Number of places: 1

 

19. Analysis, modeling and optimization of modern wireless networks.

Supervisor: dr hab. inż. Katarzyna Kosek-Szott

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The proposed research area is focused on the analysis/modeling/optimization of the performance of modern wireless networks. As part of the work it will be necessary to understand the current state of the art regarding wireless networks (e.g., Wi-Fi, LAA, 5G, 6G) and the known methods of analysis/modeling/optimization of these networks. Next, it will be necessary to select a specific research area, which may include such research directions as machine learning, Internet of Things (IoT), or software-defined networks (SDN). The expected result of the research should include, for example, new mechanisms/new ways of performance optimization/new mathematical or simulation models of modern wireless networks.

Research facilities: The Department of Telecommunications of the AGH University of Science and Technology is located in a new building which has a number of laboratories where the research can be conducted. The laboratories are equipped with new computers and different kinds of network devices (including Wi-Fi, IoT, and LTE equipment). The proposed supervisor has experience in conducting simulation research as well as using real and virtual network devices. Additionally, she has experience in modeling wireless networks using Mathematica and Matlab (both environments are available at the Department of Telecommunications). She also has experience in obtaining and managing research grants and has numerous international contacts with renowned experts in the field of wireless networks.

Number of places: 1

 

20. Quality Assessment for Computer Vision Applications.

Supervisor: dr hab. inż. Mikołaj Leszczuk

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The research issue is to study the quality of video used for recognition tasks and task-based multimedia applications. The background of the research topic is as follows: Users of video to perform tasks require sufficient video quality to recognize the information needed for their application. Therefore, the primary measure of video quality in these applications is the success rate of these tasks (such as recognition), which is referred to as visual intelligibility or acuity. One of the major causes of the reduction of optical clarity is loss of data, through various forms of compression. Additionally, the characteristics of the scene being captured have a direct effect on visual intelligibility and on the performance of a compression operation-specifically, the size of the target of interest, the lighting conditions, and the temporal complexity of the scene. The research will depend on performing a series of tests to study the effects and interactions of compression and scene characteristics. An additional goal will be to test existing or develop new objective measurements that will predict the results of the subjective tests of visual intelligibility.

Research facilities: The Department of Telecommunications is situated in two buildings with access to Polish Grid Infrastructure (PL-Grid) and other servers and infrastructure. Many software tools, including MATLAB, will be at our service throughout the execution of the research. Our dedicated Multimedia User quality of eXperience Laboratory is equipped with PCs. Also, the equipment in the laboratory includes cameras, speakers, headphones. The lab also includes VR headset platforms, many mobile/desktop 3D displays and TV sets from different manufacturers, for researching the field of three-dimensional image processing.

Number of places: 1

 

21. Artificial intelligence methods in detection of security incidents.

Supervisor: dr hab. inż. Bartłomiej Śnieżyński

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Nowadays, the cybersecurity problem becomes more important along with the increase of the responsibility of devices connected to the Internet. At the same time, there is also a dynamic development of artificial intelligence methods, especially the machine learning ones. As a result, it is natural to apply those methods in many areas of cybersecurity. Artificial intelligence techniques are already applied in the detection of security incidents, for example in intrusion detection systems (IDS), Security Information and Event Management systems (SIEM), malware detection and so on. There is still a need for an improvement of currently achieved results and for an adaptation to fast changing threats. One of possible directions of research is also application artificial intelligence methods in looking for potential vulnerabilities which can lead to security incidents.

Research facilities: The scientific unit and the promoter have knowledge about creation, development and application of artificial intelligence methods in various fields. Therefore, the proper scientific background is provided. The topic of cybersecurity is also present in the scientific unit, as Cybersecurity Centre of AGH is embedded as a part of Department of Computer Science. There is also a possibility to use computing power of ACC Cyfronet AGH, if such a need arises. Application for research grant is planned.

Number of places: 1

 

22. Optimization of Dependable Network Infrastructures for Distributed Computing.

Supervisor: dr hab. inż. Piotr Chołda

Auxiliary supervisor: dr inż. Artur Lasoń

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Network-supported centralized infrastructures (e.g., centralized clouds) supporting the processing and storage of information are very attractive nowadays. However, they raise considerable quality of service issues (e.g., delays). Edge or fog distributed processing can help resolve a number of these problems. However, this type of distibuted computing needs new networking approaches to provide dependability (i.e., security and reliability). Various supporting methods and algorithms will be studied to propose a solution satisfying various conditions necessary to successfully operate these environments. The research area includes: software-defined networking, network function virtualization, replication systems, on-board processing and networking (e.g., cars), support for vehicular mobility, optimization of dynamic resource allocation, availability modeling, etc.

Research facilities: Within the Department of Telecommunications AGH, there is a possibility of cooperation within the research group NFV & SDN, which deals with issues related to this research topic and whose members are potential supervisors. The Department of Telecommunications own a hardware infrastructure (including OpenStack infrastructure) that can support experimental research in the research field. The potential supervisors have applied for the European project (under CHIST-ERA programme) "Distributed Intelligent Vehicles Fog Networks and Applications (DIVINA)" that will focus on very research topic issues. If funding for the project is granted, the PhD student will obtain a scholarship under this financing.

Number of places: 1

 

23. Quality Metric for Video Used by First Responders.

Supervisor: dr hab. inż. Lucjan Janowski

Auxiliary supervisor: dr inż. Krzysztof Rusek

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Video quality ratings are most commonly used for systems where the viewer is watching a film for entertainment. However, numerous video systems work for usability purposes. Examples of such systems are monitoring systems, car cameras, or cameras dressed by police officers. For such systems there is also a need to create an algorithm which evaluates the quality of the video signal. The quality assessment in this case differs from the traditional one, because the usable part of the image is much more important than its proper composition. The work is focused on finding the correct solution considering deep neural networks, but not necessary limiting only to them.

Research facilities: The supervisor is strongly involved in the work of the VQEG (Video Quality Expert Group). This allows access to the relevant sequences and support from other experts working on the problem. The Telecommunications Department provides a possibility to carry out necessary computational work.

Number of places: 1

 

24. Optimization of Wireless Local Area Networks based on newest IEEE 802.11 amendments.

Supervisor: dr hab. inż. Marek Natkaniec

Auxiliary supervisor: dr inż. Janusz Gozdecki

Faculty of Computer Science, Electronics and Telecommunications

Abstract: In recent years, an extremely dynamic development of subsequent amendments of the IEEE 802.11 standard has been observed. They introduce changes both at the level of the physical layer and the medium access control layer. The increasing number of configurable parameters causes problems with obtaining the optimal and comparable values ​​of typical performance metrics (throughput, average transmission delay, jitter, frame losses) for different network configurations and topologies. This creates problems both in terms of the quality of the modern multimedia services provisioning and unfair access to a common radio channel. The aim of the research topic will be to develop various algorithms, protocols and mechanisms to improve QoS, ensure fair access to the radio channel as well as automatic configuration of selected physical and MAC layer parameters to increase overall network performance for different configurations and topologies.

Research facilities: Within the Department of Telecommunications AGH, there is a possibility of cooperation within the research group Wireless Network, which deals with issues related to this research topic and whose head is supervisor of this research topic. The Department of Telecommunications has the appropriate research tools (network simulators, software for modeling network protocols, computing servers) and access to science databases (e.g. IEEExplore, Elsevier, Wiley) as well as hardware infrastructure that can support experimental research in the field of the issue (including several dozen types of WLAN cards and bridges based on IEEE 802.11a/b/e/g/h/i/n/s/ac standards). KT AGH cooperates with many recognized European universities in the area of wireless local area networks.

Number of places: 1

 

25. Optimization of Wi-Fi or LTE networks using artificial intelligence (machine learning) or game theory algorithms.

Supervisor: dr hab. inż. Szymon Szott

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Modern wireless networks (Wi-Fi and LTE) are constantly being extended with new functionalities. The standards defining these networks (developed by IEEE and 3GPP) define new possibilities, but they do not determine how to use them. This applies to such aspects as the selection of network operation parameters (e.g., transmit power and operation frequency) and allocation of resources among users. There is an urgent need to develop methods for optimizing the operation of wireless networks working in various scenarios and based on the latest versions of the IEEE / 3GPP standards. Artificial intelligence algorithms stand out among the available methods of optimization (in particular algorithms based on machine learning). In addition, for distributed environments, game theory offers tools that allow systems to achieve optimal work points. Currently, however, there are no solutions dedicated to the latest versions of the wireless network standards. Not only research centers are interested in this subject, but also network operators and equipment manufacturers. This means that in the area described above, many interesting research hypothesis can be defined.

Research facilities: The Faculty of Computer Science, Electronics and Telecommunications has the necessary research facilities to conduct the aforementioned research: both in terms of programming tools, computing power, and equipment (high availability of Wi-Fi devices, own LTE network). After specifying the subject of the PhD thesis, it is planned to submit an application for a scientific project (NCN or NCBR). The subject matter may also be of interest to companies producing Wi-Fi or LTE equipment with which the Faculty cooperates.

Number of places: 1

 

26. ldentification of patterns and anomalies in social media.

Supervisor: dr hab. inż. Jarosław Koźlak

Auxiliary supervisor: dr inż. Anna Zygmunt

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Social media play a significant and ever-growing role in the functioning of modern society. They influence various dimensions such as business and professional development, politics, marketing, shaping social relations, development of private interests, etc. We consider it useful to better understand the behavior of users forming communities in the social media, frequent patterns and regularities, as well as identifying characteristic anomalies. In particular, we are interested in behavior patterns related to time, characterized by repeatability or stability. For such patterns, one can study their co-occurrence and predict whether appropriate patterns will be stili present within the time period considered. We use approaches based on the social and complex networks analysis, data mining, machine learning and artificial intelligence.

Research facilities: The Department of Computer Science provides the appropriate research facilities and equipment needed to carry out the research. In addition , if there is such a need, it is possible to get access to supercomputers in the Cyfronet computing center, which provides one of the most powerful computation capabilities in our country.

Number of places: 1

 

27. Adaptive traffic engineering in IP and optical Software Defined Networks.

Supervisor: prof. dr hab. inż. Zdzisław Papir

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The topic of research are adaptive and multipath traffic engineering algorithms, dedicated for usage by network operators in wide-area IP and optical networks. In particular, focus is put on adaptive traffic engineering in SDN IP networks and traffic engineering in multilayer (IP+optical) networks. Existing FAMTAR (Flow-Aware Multi-Topology Adaptive Routing) and AHB (Automatic Hidden Bypasses) mechanisms will be the starting points of the research. The possibility of their realization in SDN networks and joint usage for traffic engineering in multilayer network will be researched.

Research facilities: Department of Telecommunications owns hardware required for research on the topic. PhD student will have an access to computer networks lab and optical networks lab. Department owns several multiprocessing servers, which can be used for simulations and emulation of virtual networks with implemented algorithms. Apart from that, PhD student will have a possibility to use PL-GRID computing infrastructure.

Number of places: 1

 

28. Agent-based modeling and simulation.

Supervisor: prof. dr hab. inż. Grzegorz Dobrowolski

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Research in agent-based modeling and simulation in specific application domains sharing the characteristics of complex (large) systems. This approach assumes modeling of autonomous subsystems using artificial intelligence methods. Such studies will be conducted in two general strategies: explaining the functioning of real systems and studies on the development of real systems built - in a broad sense - by a human being. Models and intentional simulation studies are being built. Examples of domains are: industry management, transport.

Research facilities:

Number of places: 1

 

29. Machine learning for cybersecurity.

Supervisor: dr hab. inż. Marcin Niemiec

Auxiliary supervisor:

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The research will be focused around the development and use of machine learning algorithms for the digital data protection. One of the main research directions will be the design and application of machine learning and statistical methods to analyze data generated by network devices/systems. Such solutions can be used to detect anomalies and malware as well as security threats and vulnerabilities in protected systems/applications. These studies should lead to solutions in which, it will be possible to predict/detect network attacks or malware campaigns, based on the analysis of large data sets from many sources. Another research direction will be focused on cryptographic systems based on artificial neural networks. The research on the use of machine learning in cryptography will focus on the analysis of the strengths and weaknesses of this solution, as well as on the investigation of potential attack vectors. This research direction will not be limited only to the analysis of classical cryptographic systems, but also modern solutions based on quantum techniques (e.g. quantum cryptography with error correction algorithms) will be subjected to analysis.

Research facilities: Department of Telecommunications has the equipment and software needed for research within this research topic. PhD student will have an access to network laboratories, including the network security laboratory. Department has computing servers where a PhD student will be able to simulate the proposed algorithms and mechanisms. Additionally, a PhD student may be involved to work in international H2020 ECHO project (European network of Cybersecurity centers and hub for innovation and Operations).

Number of places: 1

 

30. Analysis of publicly listed companies by using artificial intelligence systems.

Supervisor: dr hab. Andrzej Bielecki, prof. AGH

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The doctoral thesis will aim to verify the effectiveness of artificial intelligence methods in the analysis of publicly listed companies. In particular, an attempt will be made to test the following research hypotheses:

H1: Artificial intelligence and machine learning algorithms allow identification of entities manipulating the financial report.

H2: Artificial intelligence and machine learning algorithms allow identification of companies committing economic crimes.

H3: Artificial intelligence and machine learning algorithms predict economic slumps and economic crises.

H4: Artificial intelligence and machine learning algorithms allow to identify forms of unfair competition (for example, intentional lowering of prices below costs).

H5:Artificial intelligence and machine learning algorithms (including those based on Benford law) allow the identification of entities that evade taxation.

Hypotheses are going to be verified with machine learning and artificial intelligence algorithms. Asa part of the research, a comparative analysis of existing methods will be made to assess their effectiveness as well as the speed of calculations. In the further part of the study, an attempt will be made to develop existing tools to improve the efficiency and speed of existing methods.

Research facilities: In order to conduct the declared studies the access to Reuters Datastream, Orbis, Jstor databases is needed. The Chair of Applied Computer Science will ensure the access.

Number of places: 1

 

31. Methodically Unified Procedures for Outliers Detection, Clustering, and Classification in Conditional Approach.

Supervisor: prof. dr hab. inż. Piotr Kulczycki

Faculty of Physics and Applied Computer Science

Abstract: The subject of the studies are procedures for the identification of atypical elements (outliers), clustering, and classification for the conditional case, i.e. when distribution characteristics of the dataset are dependent on quantities metrologically available (e.g. current temperature), which in practice efficiently allows the model to be made used more precise and up-to-date. In order to solve the issue thus formulated. nonparametric estimation methods will be used, which frees the procedures under research from the distribution in the investigated dataset. Elements of computational intelligence - fuzzy logic (including intuitionistic fuzzy sets) and genetic algorithms - will be applied in particular aspects. The results obtained will be illustrated and tested using synthetic data and benchmarks, and also research in environmental engineering conducted at the Faculty of Physics and Applied Computer Science at AGH, as well as - optionally - in a domain proposed by the Ph.D.-student. The unconditional case was successfully investigated out, verified, and applied [1-4] in cooperation with former Ph.D.-students Małgorzata Charytanowicz, D.Sc., Piotr A. Kowalski, D.Sc., Damian Kruszewski, Ph.D., and Szymonem Łukasik, Ph.D. Mathematical predispositions and programing ability are required from the Ph.D.-student. Contact person: Prof. Piotr Kulczycki, e-mail: kulczycki@agh.edu.pl

Research facilities: The Faculty will provide complete IT facilities and other infrastructure for the realization of the planned research.

Number of places: 1

 

32. Synthesis of moving switching curve based on nonstationary streaming data.

Supervisor: prof. dr hab. inż. Piotr Kulczycki

Auxiliary supervisor:

Faculty of Physics and Applied Computer Science

Abstract: This topic from the streaming data analysis domain, contains also significant elements of control engineering. The task relies on the synthesis of switching surface occurring among others in variable structure control (also in sliding and robust control), in an adaptive form, i.e. making its time variable shape and position dependent on information concerning the object, incoming successively in the form of a data stream. The research will be carried out for selected nonlinear systems, also described by differential equations with discontinuous right-hand side. Conditional approach will be also investigated, where the characteristics of the object are significantly dependent on quantities whose current value is available metrologically, which often allows the inference process to be considerably more precise. It will be possible (although not necessary) to construct a laboratory setup for the empirical verification of the results. Mathematical predispositions and programing ability are required from the Ph.D.-student. Contact person: Prof. Piotr Kulczycki, e-mail: kulczycki@agh.edu.pl

Research facilities: The Faculty will provide complete facilities for the realization of the planned research.

Number of places: 1

 

33. Sensitivity analysis of convolutional neural networks.

Supervisor: dr hab. inż. Piotr A. Kowalski

Auxiliary supervisor:

Faculty of Physics and Applied Computer Science

Abstract: The subject of the research will be the innovative development Sensitivity Analysis (SA) for Deep Neural Networks in particular Convolutional Neural Network. The main task of the SA algorithms will be to reduce the individual components of deep neural networks, aimed at examining both the impact (substantiality) of individual components and the simplification of the structure. SA approaches can be categorized into the following two groups: Local Sensitivity Analysis (LSA) and Global Sensitivity Analysis (GSA). LSA explores the changes of model response by varying one parameter while keeping the other ones constant. The simplest and most common LSA approach is based on partial derivatives of the output functions with respect to the input parameters. In GSA, the influence on models’ outputs can be evaluated using regression methods, screening approaches and the variance-based techniques, e.g., Sobol, the Fourier amplitude sensitivity test (FAST) or the extended FAST (EFAST). In this investigation, the following approaches for the structure simplification of the considered network will be proposed: (i) an algorithm reducing solely the number of input neurons, (ii) an algorithm decreasing solely the number of convolutional neuros, and (iii) an algorithm removing neurons in fully connected layers and (iv) finally all above procedures will be merge for removing input and convolutional and fully connected neurons simultaneously.

Number of places: 1

 

 

34. Solving interval parametric linear systems

Supervisor: dr hab. Iwona Skalna

Faculty of Management

Abstract: Solving interval parametric systems is a very important issue in many fields of science and technology. The work will aim to develop new methods for solving such systems using various methods of bounding functions on an interval, including quadratic forms, Taylor series, Bernstein polynomials.

Number of places: 3

 

35. Joint Treatment of Imprecision and Randomness in the Appraisal of the Effectiveness and Risk of Investment Projects

Supervisor: dr hab. Iwona Skalna

Auxiliary supervisor: dr Bartłomiej Gaweł

Faculty of Management

Abstract: Every company must make investment decisions, often under conditions of uncertainty. The latter may arise from a wide range of sources at different points of time in a project lifecycle. The variety of sources causes that different types of uncertainty can occur simultaneously. Quantification of risk associated with decision-making processes in such circumstances is a very complex task. The work will aim to develop new methods for evaluating the effectiveness and risk of investment projects in the simultaneous presence of both imprecise and stochastic uncertainty is proposed.

Number of places: 2

 

36. Nanocommunications

Supervisor: prof. dr hab. Inż. Andrzej Jajszczyk Auxiliary supervisor: dr inż. Paweł Kułakowski

Department of Telecommunications,Faculty of Computer Science, Electronics and Telecommunications

Abstract: Nanocommunications is a relatively new research area focused on communications between nanomachines. The research conducted in the Department of Telecommunications of AGH is carried in two fields: (a) electromagnetic communication (in THz band) between micromachines built of, e.g., graphene, (b) molecular communication using bio-inspired mechanisms like information transfer of data coded in DNA chains or the phenomenon of FRET. The doctoral thesis would be about carrying both experimental and simulation-based research on feasibility of nanocommunications for bio-medical purposes (like nanomachines communicating inside of a human body).

Research facilities: Research in the area of nanocommunications is realized using the PLGrid Infrastructure (for simulations and large-scale calculations) and scientific databases like IEEExplore, Wiley and Elsevier. The Department of Telecommunications cooperates with other research universities like Polytechnic University of Cartagena and University of Lisbon and, located in Krakow, Collegium Medicum Jagiellonian University and the Department of Medical Physics and Biophysics of AGH, carrying joint research, also measurements using, e.g., a confocal microscope. The research on nanocommunications realized at AGH is also a part of large European research projects (COST) like IC1004 or IRACON. There are possibilities of research stays abroad realized in the frame of these projects.
Number of places:
1