Deutsch Intern
    Data Science Chair

    Theses and Practica

    The following pages show open topics for practica, bachelor and master theses, sorted by application fields. We will be happy to provide information on further open topics upon request. Topics can be completed either in English or German.

    Open Topics

    05/15/2025 | Physics & Deep Learning, Ecosystems

    Autoregressive Deep Learning Earth System Models

    Interested in AI and Climate Science? Simulating the Earth’s climate with Earth System Models (ESMs) is essential for understanding and predicting climate change — past, present, and future. These physics-based simulations are… more
    05/08/2025 | Security

    LLM-based Intelligent Literature Review Assistant

    Develop a semi-automated assistant for structured literature reviews, combining LLMs, citation graph search, and spreadsheet-based curation. This project builds on an existing prototype with basic PDF parsing, citation crawling,… more
    05/07/2025 |

    Using Explainable AI to Uncover Gene Relevance in scRNA-seq Data

    Single-cell RNA sequencing (scRNA-seq) provides detailed insights into gene expression at the cellular level. In this thesis, we use explainable AI (XAI) methods to investigate which genes most influence the predictions of a… more
    05/07/2025 |

    Assessing the Suitability of (Zero-Inflated) Negative Binomial Distributions for scRNA-seq Gene Expression Data

    Single-cell RNA sequencing (scRNA-seq) data is highly sparse, with many zero values that pose challenges for statistical modeling. Two suitable distributions to use are the negative binomial (NB) and the zero-inflated negative… more
    05/06/2025 | Security

    Elastic Stack Integration for Cyber Security Simulation Logging

    Adapt the existing CIDDS simulation environment so that system and network logs are forwarded directly into the Elastic Stack. Evaluate logging tools (e.g., Winlogbeat, Filebeat, Auditbeat) and configure them for simulated… more
    05/06/2025 | Security

    Automating Attack Simulations with MITRE ATT&CK

    Extend an existing simulation environment to execute reproducible cyber-attack scenarios using the MITRE ATT&CK framework and procedures for example from Aomic Red Team. The focus is on automation of attack step execution and… more
    05/06/2025 | Security

    XAI Benchmarking for Cyber Security

    Investigate how to derive explanation-relevant features from simulated SIEM logs by analyzing known attack scenarios. Use publicly available annotated attack logs (e.g., from securitydatasets.com) to identify explanation-relevant… more
    05/06/2025 | Security

    Domain-Specific Representation-Learning for Cyber Security

    Develop and evaluate representation learning methods on structured security log data. The goal is to capture domain-specific patterns from benign and attack behavior that can later support anomaly detection and explainability in… more
    04/30/2025 | Security

    Role-Based User Behavior Simulation in Enterprise IT Environments

    Develop lightweight simulation agents that mimic realistic user behavior (e.g., office staff, developers, admins) in a controlled lab setup. Agents should execute atomic tasks such as browsing, file access, or command usage, with… more
    03/03/2025 | Recommender Systems

    Knowledge-Graph based Recommender Systems for e-Learning on CaseTrain data

    deep learning for recommendation
    The goal is to development and improve recommender systems for e-learning based on data from the CaseTrain system. more
    02/17/2025 | Natural Language Processing

    Exploring and Evaluating Various Domain Adaptation Techniques

    Domain adaptation involves modifying a model, originally trained on general data, to perform effectively in more specialized fields such as medicine, politics, law or instruction following. more
    01/30/2025 | Security

    Agent-based Machine Learning for Offensive and Defensive Cyber Security

    Automatisierte Agenten bieten ein enormes Potenzial für offensive und defensive Cybersecurity. Von der Informationsbeschaffung über Angriffe bis hin zur Erkennung und Abwehr von Angriffen können sie eine Vielzahl komplexer… more
    01/30/2025 | Security

    Knowledge-based Machine Learning for Proactive Cyber Threat Modeling and Detection Using Ontologies and Knowledge Graphs

    Die Erkennung und Modellierung von Cyberangriffspfaden erfordert eine automatisierte Analyse von Bedrohungen. Frameworks wie MITRE ATT&CK oder D3FEND bieten umfangreiche Wissensbasen, werden jedoch oft isoliert genutzt. more
    01/29/2025 | Physics & Deep Learning, Ecosystems

    Enhancing Regional Climate Projections with Deep Learning

    Global climate models play a vital role in predicting future climate conditions, such as temperature or precipitation. However, their coarse spatial resolution (typically 50–100 km) limits their ability to capture regional climate… more
    01/22/2025 | Natural Language Processing

    Klassifikation von Romanen anhand von Zeitreihen

    Romane gehören zu den Textarten, die trotz riesiger Fortschritte in den letzten Jahren die NLP immer noch vor Herausforderungen stellen: Selbst die neuesten Modelle sind nicht in der Lage, komplette Romantexte am Stück einzulesen… more
    01/22/2025 | Physics & Deep Learning, Ecosystems

    Spherical Embeddings or Spherical Fourier Neural Operators

    Spherical Harmonics by Inigo.quilez, licensed under CC BY-SA 3.0, via Wikimedia Commons, (https://commons.wikimedia.org/wiki/File:Spherical_Harmonics.png)
    Although our planet is a sphere and therefor all weather and climate processes operate on sphere, Deep Learning Weather Models very commonly operate on different rectangular projections of the sphere. This introduces increasingly… more
    11/13/2024 | Physics & Deep Learning, Ecosystems

    Meteorological Downscaling Using Remote Sensing Data

    Global numerical climate models project meteorological quantities such as temperature, humidity, pressure, etc. decades into the future under the assumption of different emission scenarios. However, due to computational… more