Open Topics for Bachelor and Master Theses

The following lists include open research topics available for bachelor and master theses as well as practical projects. Please note that these lists do not exhaustively list all topics. Members of the chair often offer additional topics within their area of interest, so feel free to contact them directly at their personal web page.

Regarding general questions please contact Stefan Geißler.

You can easily view a summary and a detailed description of the topics listed here by clicking the Abstract link below the respective entry. The contact details of the respective supervisor are also made available there.

Open Bachelor Theses

Supervisor: Dietz, K.: Agree to disagree? Benchmarking outlier detection algorithms for network security, (2022).
- [ Abstract ]
- [ Download ]
The detection of outliers in network traffic is an omnipresent topic for network security, e.g., to detect malicious activity or unknown devices and apps in the network, that deviate from the normal network activity. The goal of this thesis is to compare a myriad of different detection algorithms and compare them not only regarding their accuracy and detection time, but also with regards to the false negatives and false positives, i.e., investigate the outliers they disagreed on, which were falsely detected etc., as this may be a reason to include more than one algorithm, e.g., in a monitoring system. For more information, read the linked presentation slide and/or message katharina.dietz@uni-wuerzburg.de.
Supervisor: Dietz, K.: Certainly uncertain: Simulative evaluation of (in)confident machine learning for user-based active learning, (2022).
- [ Abstract ]
- [ Download ]
Machine learning is becoming more and more prevalent in networking, e.g, network monitoring and managament. Though, expert knowledge is always valuable, sometimes irreplaceable. As the machine learning model may not always be confident, one idea is to relay such inconfident decisions to an expert, i.e., the admin, which can decide what to do with this decision and potentially relabel it, called user-based active learning. The goal of this thesis is to perform a simulative analysis regarding the influence of various parameters, such as the confidence threshold (i.e., when to relay the information) or the processing time and qualification level of the admin (i.e., a fast but not always correct expert, slow but accurate etc.). For more information, read the linked presentation slide and/or message katharina.dietz@uni-wuerzburg.de.
Supervisor: Pruckner, M.: Modeling and simulation of sustainable local energy systems, (2022).
- [ Abstract ]
Due to the raising costs for energy and CO2-emission, more and more small companies are interested in new decentralized technologies for CO2-neutral local energy production and storage. For instance, companies with high electricity and/or process-heat demand are interested in photovoltaic and/or solar thermal systems as well as combined heat and power systems (e.g., heat pump - KWK). We want to go one step further: We want to design a new type of system, that will harvest the power of the sun (photovoltaic and/or solar thermal) and provide electricity, warmth and cold (combined heat, power and cooling - KWKK) to use in manufacturing processes. The supply of the energy forms shall be local for self-sufficiency as well as CO2-neutral. The system should be flexible, so that the user can determine which of the three energy forms he/she needs most. Storing electricity will be realized in accumulators while cold and warmth will be stored in insulated water tanks. In this bachelor thesis, a data-driven simulation model should be developed, based on a use case. Based on real and open data different energy system components should be implemented. The model should be used to evaluate different scenarios for sustainable energy supply by varying different parameters such as the size of the photovoltaic system or the capacity of the battery energy storage. Finally, an evaluation of cost savings and CO2 reduction potentials should be revealed.
Supervisor: Seufert, M.: Models for Group Communication Behavior under Mobility, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de --- Abstract: In this work, a literature study shall be conducted to survey existing models for mobility and communication behavior, i.e., how people communicate while they are moving (e.g., walking, riding a bus). Selected models shall be implemented and implications of mobility on communication behavior shall be simulatively derived. --- Inhalt: In dieser Arbeit soll eine Literaturstudie durchgeführt werden, um einen Überblick über existierende Modelle für Mobilität und Kommunikationsverhalten zu erhalten, d.h. wie Menschen kommunizieren während sie unterwegs sind (z.B. laufen oder Bus fahren). Ausgewählte Modelle sollen implementiert werden und die Auswirkungen von Mobilität auf das Kommunikationsverhalten sollen simulativ abgeleitet werden.
Supervisor: Dietz, K.: Comparing apples and oranges? Assessment and selection of features for transfer learning for network security, (2022).
- [ Abstract ]
- [ Download ]
The acquisition of adequate data is a crucial concern for machine learning and especially network security. One approach to circumvent this problem is transfer learning, where data from another - but similar - domain is reused to solve the new problem. Though, even if the data is similar, it may not be exactly the same, i.e., a perfect fit. Thus, the goal of this thesis is to evaluate compatible data sets on a statistical basis and subsequently, test transfer learning approaches via machine learning. For more information, read the linked presentation slide and/or message katharina.dietz@uni-wuerzburg.de.

Open Master Theses

Supervisor: Pruckner, M.: Passenger transport with minibuses, (2023).
- [ Abstract ]
- [ Download ]
Minibuses are already being used to transport passengers on public transport lines with low demand. Minibus lines are often operated like normal bus lines (based on schedules) or like shared cabs (transport requests must be announced in advance). The question arises whether (and under which conditions) there are better and more efficient strategies for operating a minibus line. In this context, there are various questions that can be addressed in the context of a final thesis. An interesting task would be the creation of a concrete test instance based on already operating or planned minibus lines, the development of a simulation that allows to test and compare different strategies, and the development and (experimental and/or theoretical) analysis of strategies.
Supervisor: Pruckner, M.: Charging planning for electric buses, (2023).
- [ Abstract ]
- [ Download ]
If a transit agency replaces part of its conventional bus fleet with electric buses, the allocation of buses to trips must be rethought to allow for sufficient charging time between trips for the electric buses. Factors such as fluctuating electricity prices, temperature-dependent energy consumption should also be considered. In preliminary work, this optimization problem was modeled as a mixed-integer linear program, which in principle can be used for potential studies ('How much energy could have been maximally saved for a specific time period by using electric buses?'), but in its current form can only be used for very short time periods due to its long runtime. The task of this thesis is to find an approach, based on the present model, to solve the problem for longer time periods (such as one year). For this, different (exact or heuristic) approaches are possible, for example a (partial) re-formulation of the model or a 'rolling horizon' approach (where the model is solved step by step for shorter, overlapping time periods. It would further be interesting to adapt the model for use in operational planning, where future energy prices and temperature conditions are subject to uncertainty.
Supervisor: Seufert, M.: Design and Performance Evaluation of a Decomposition-based Approach to Network Traffic Classification, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de Title: Design and Performance Evaluation of a Decomposition-based Approach to Network Traffic Classification Abstract: Traffic classification, i.e., the identification of the application type based on its network traffic, is an important prerequisite to application-specific network monitoring. In this work, a decomposition based approach to network traffic classification shall be designed and implemented. Here, decomposition means that the network traffic shall be split into meaningful parts, and for each part, relevant patterns shall be identified. The performance of the implemented approach shall be compared to other state-of-the-art approaches for network traffic classification. Titel: Design und Leistungsbewertung eines zerlegungsbasierten Ansatzes für die Klassifizierung von Netzverkehr Inhalt: Die Klassifizierung von Netzverkehr, also die Identifizierung des Typs der Anwendung basierend auf ihrem Netzverkehr, ist eine wichtige Voraussetzung für anwendungsspezifisches Netzmonitoring. In dieser Arbeit soll ein zerlegungsbasierter Ansatz für die Klassifizierung von Netzverkehr entworfen und implementiert werden. Zerlegung bedeutet hier, dass der Netzverker in sinnvolle Teile aufgespaltet werden soll und für jeden Teil relevante Muster identifiziert werden sollen. Die Leistungsfähigkeit des implementierten Ansatzes soll mit anderen State-of-the-Art-Ansätzen zur Klassifizierung von Netzverkehr verglichen werden.
Supervisor: Strobel, L.: Electrification of public transport in European towns - A case study of Bayreuth, (2022).
- [ Abstract ]
The electrification of public transport is a major milestone towards carbon neutrality. This process is well on its way in large urban centers, but in smaller towns, the vehicle inventory still primarily consists of fossil-fuel-powered buses. Significant uncertainty about the economic and practical feasibility of full-electric fleets is prevalent among stakeholders. At the same time, public transport is uniquely qualified for smart-charging use cases due to the centralized control of the transport authority and the highly predictable nature of driving patterns. This thesis aims to evaluate the feasibility of switching to an all-electric bus fleet in Bayreuth and the resulting smart-charging revenue potential. This work is part of ESM-Regio, a project in partnership with Stadtwerke Bayreuth, and real-world data is available. The thesis includes: - Literature research - Analysis of the available data - Development of electric bus model - Development of a charging optimization model - Scenario analysis Requirements: - Knowledge of any programming language Helpful: - Experience with data analysis - Experience with mathematical optimization (We can do a crash course)
Supervisor: Seufert, M.: Design and Comparison of Heuristic- and DL-based Estimators of QoE Factors from Visual Recordings, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de --- Abstract: A heuristic- and/or deep-learning-based software shall be implemented and evaluated, which allows to detect QoE factors from a visual recording of the client's screen, such as video resolution, frame rate, or stalling. --- Inhalt: Eine auf Heuristiken und/oder Deep Learning basierte Software soll implementiert und evaluiert werden, die es erlaubt QoE Faktoren aus einer Videoaufnahme des Bildschirm des Clients zu erkennen, wie z.B. Videoauflösung, Framerate oder Stalling.
Supervisor: Seufert, M.: Deep Learning-based Transfer Learning for QoE Monitoring of Encrypted Video Streaming Traffic, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de Title: Transfer Learning for QoE Monitoring of Encrypted Video Streaming Traffic Abstract: This work is focused on Quality of Experience (QoE) monitoring of encrypted video streaming traffic. Here, machine learning based methods are used to estimate QoE metrics based on features or embeddings, which are extracted from the encrypted traffic. While these machine learning models achieve very good performance for a specific training data set, they typically show a worse performance when applied to different data sets. Thus, in this thesis, deep learning based transfer learning approaches shall be tested and/or developed, which allow the existing models to generalize also over heterogeneous data sets. The performance of the transfer learning approaches shall be evaluated and a methodology for generalizing machine learning based QoE monitoring models shall be developed. Titel: Transferlernen bei der QoE Messung auf verschlüsseltem Videostreamingverkehr Inhalt: Diese Arbeit beschäftigt sich mit der QoE Messung auf verschlüsseltem Videostreamingverkehr. Hierbei werden Methoden des maschinellen Lernens (ML) eingesetzt um QoE-Metriken zu schätzen basierend auf Features oder Embeddings, die aus dem verschlüsselten Verkehr extrahiert werden. Während diese ML-Modelle sehr gute Leistung für spezifische Datensätze zeigen, haben sie typischerweise eine schlechtere Leistung wenn sie auf andere Datensätze angewandt werden. Deshalb sollen in dieser Arbeit Deep Learning-basierte Transferlernverfahren getestet und/oder entwickelt werden, die es erlauben, dass die existierenden Modelle auch für heterogene Datensätze verallgemeinert werden können. Die Leistung der Transferlernansätze soll bewertet werden und eine Methodik für die Generalisierung von ML-Modellen für QoE Messungen soll entwickelt werden.

Available Student Projects

Supervisor: Seufert, M.: Data Augmentation of Network Traffic Traces, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de --- Abstract: In this project, several approaches to data augmentation for network traffic traces shall be implemented and evaluated. The goal is to generate synthetic traffic traces, which have almost identical properties to real traffic traces. This allows to reduce the amount of network traffic traces needed to train ML-based models. --- Inhalt: In diesem Praktikum sollen verschiedene Ansätze zur Datenanreicherung von Aufzeichnungen von Netzverkehr implementiert und evaluiert werden. Das Ziel ist es synthetische Verkehrsaufzeichnungen zu erzeugen, die beinahe identische Eigenschaften haben wie echte Verkehrsaufzeichnungen. Dies ermöglicht es die benötigte Menge von Aufzeichnungen von Netzverkehr beim Training von ML-basierten Modellen zu reduzieren.
Supervisor: Seufert, M.: Identifying User Behavior from Encrypted Video Streaming Traffic, (2022).
- [ Abstract ]
Contact/Kontakt: michael.seufert@uni-wuerzburg.de Title: Identifying User Behavior from Encrypted Video Streaming Traffic Abstract: In this work, existing PC- and app-based measurement frameworks have to be modified to conduct video streaming measurements with emulated user interactions. Application and network layer streaming information shall be extracted and analyzed to identify the emulated user interactions from the encrypted streaming traffic. Titel: Indentifizierung von Nutzerverhalten aus verschlüsseltem Videostreaming-Netzwerkverkehr Inhalt: In dieser Arbeit sollen existierende PC- und App-basierte Messframeworks modifiziert werden, um Messungen von Videostreaming durchzuführen und Nutzerinteraktionen zu emulieren. Streaminginformationen sollen auf Applikations- und Netzwerkschicht extrahiert werden, um die emulierten Nutzerinteraktionen aus dem verschlüsselten Videostreaming-Netzwerkverkehr zu identifizieren.
Supervisor: Poignée, F.: Simulating Group-based Communication with Omnet++, (2021).
Supervisor: Geissler, S.: Design and Implementation of a WebApp for the Performance Evaluation of VNF Chains, (2021).
- [ Abstract ]
Goal of this work is the design and implementation of a web-based application that allows the graphical configuration of service function chains (SFC). An underlying backend (written in R) should then perform detailed performance evaluation (discrete-time modeling, simulation) of the designed SFC and generate several key performance indicators of the system.
Supervisor: Loh, F.: Studying mobility in a LoRaWAN Network, (2021).