Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
Thesis. Contact: Ronny Bodach
More about this thesis
Description: Development of a realistic and practice-oriented scenario for a cyber crisis simulation that targets current threats and challenges in the field of cyber security, support in creating training materials for ‘Cyber Crisis Management - Experiencing Crisis Simulation Live’ and evaluation tools to assess participants during and after the simulation.
Bachelor thesis. Contact:
Amon Bentke
More about this thesis
Description: Super recognizers are people with extraordinary face recognition abilities that enable them to identify faces with a remarkably high degree of accuracy, even after long periods of time or under challenging conditions. In law enforcement, super recognizers play a crucial role by helping to identify suspects or missing persons. In contrast, artificial intelligence (AI) has made significant advances in facial recognition in recent years. Using machine learning and neural networks, AI can analyze large amounts of data and recognize patterns in faces to automate identification. This technology is increasingly being used in various applications, from surveillance to access control, and has proven to be highly efficient. Despite the progress made in both human super recognizers and AI, the question remains as to where the limits of human abilities lie and to what extent the performance of AI can exceed or complement them. Can the abilities of super recognizers also be used in other ways? How do super recognizers compare to artificial intelligence when it comes to accurately recognizing faces?
Main tasks: Evaluation and discussion of a test specially designed for Super-Recognizer in Saxony and carried out on September 25, 2024, evaluation of the data set on which the test is based using modern methods of artificial intelligence, and discussion of the results, comparison between Super-Recognizer and AI, further development of the Super-Recognizer test if necessary.
Requirements: Experience with quantitative analysis methods and basic knowledge of statistics, basic programming skills in Python, initiative, creative ideas and an interest in trying out and learning something new.
Note: The aim is to publish your work in a scientific journal at the end.
Bachelor thesis. Contact:
Svenja Preuß
More about this thesis
Tasks: Analyze MasterLock Vault Home App, examine the app in a virtual environment, collect forensic traces, classify / categorize forensic traces according to relevance
Requirements: Basic knowledge of Linux / Android, structured and methodical approach, organized literature research
Internship / Bachelor thesis. Contact:
Felix Fischer
More about this thesis
Tasks: Analyze Abus One app, examine the app in a virtual environment, collect forensic traces, classify / categorize forensic traces according to relevance
Requirements: Basic knowledge of Linux / Android, structured and methodical approach, organized literature research
Internship / Bachelor thesis. Contact:
Felix Fischer
More about this thesis
Tasks: Analyzing the Jura Operating Experience app, examining the app in a virtual environment, analyzing network traffic to the coffee machine, collecting forensic traces, classifying / categorizing forensic traces according to relevance
Requirements: Basic knowledge of Linux / Android, basic network knowledge, structured and methodical approach, organized literature research
Internship / Bachelor thesis. Contact:
Felix Fischer
More about this thesis
Tasks: Analyze Bosch Smart Camera app, examine the app in a virtual environment, analyze forensic evidence, classify / categorize forensic evidence according to relevance
Requirements: Basic knowledge of Linux / Android, Structured and methodical approach, Organized literature research, Understanding of network communication, Structured and methodical approach, Organized literature research
Bachelor thesis / Master thesis. Contact:
Felix Fischer
More about this thesis
Tasks: Recording network traffic, analyzing network traffic, identifying weak points, documenting transmitted information, Bosch contact sensor, SwitchBot contact sensor
Requirements: Understanding of network communication, structured and methodical approach, organized literature research
Internship / Bachelor thesis. Contact:
Felix Fischer
More about this thesis
Tasks: Provision of software libraries C++, provision of hD2 microcontroller board, uniformly structure software libraries, establish compatibility of software libraries with each other, programming a demo program, developing new software libraries, improving code quality
Requirements: Programming knowledge (C++ is used), structured and methodical approach, organized literature research
Internship. Contact:
Felix Fischer
More about this thesis
Tasks: Analyze Bosch Smart Camera App, examine the app in a virtual environment, examine the app on rooted Pixel 7a, analyze forensic traces, classify forensic traces by relevance
Requirements: Basic knowledge of Linux / Android, structured and methodical approach, organized literature research
Bachelor thesis / Master thesis. Contact:
Felix Fischer
More about this thesis
When analysing forensic data, it is often crucial to incorporate the investigators' knowledge (e.g., the criminal liability of actions, the circumstances of the crime). Many automated methods for analysing data require a formal description of this knowledge in the form of knowledge representations. Such knowledge representations range from familiar mind maps to semantic networks (e.g. WordNet) and ontologies. To determine which knowledge representations are particularly promising for which use forensics cases, a systematic literature review should be conducted. In particular, this is intended to demonstrate the use of knowledge representation in forensics and crime prevention.
Bachelor's thesis. Contact:
Michael Spranger
More about this thesis
When investigators analyse forensic communication data, they often already have an assumption as to which case-relevant topics were discussed. To test this hypothesis, they look for clues to the relevant topics in the immense volume of messages. Semi-supervised topic modelling can be used for this purpose. The semi-supervised algorithms are given characteristic words to describe each case-relevant topic. However, the success of semi-supervised topic modelling depends heavily on selecting these topic-specific words. The aim is therefore to apply semi-supervised topic modelling iteratively and identify new case-relevant words on the basis of the topics found.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...), investigators often expect specific topics to have been discussed. Their goal is, therefore, to find clues or evidence for these case-relevant topics. There is potential here in semi-supervised topic modelling. This allows knowledge about expected topics to be considered by passing a few characteristic words as input for each topic. However, finding suitable words to describe case-relevant topics is often tricky. The aim is therefore to use Large Language Models (LLMs) such as ChatGPT to identify more meaningful synonyms for user-defined initial terms and thus improve the results of semi-supervised topic modelling.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...), investigators often start from a specific hypothesis, for example, that a particular offence has been committed. Accordingly, they expect topics related to this offence to be discussed in the chats and want to find evidence of this. Semi-supervised topic modelling can be used for this purpose. This allows knowledge about expected topics to be considered by passing a few characteristic words as input for each topic. However, finding suitable words to describe case-relevant topics is often tricky. The aim is, therefore, to use word embeddings to identify more meaningful, similar words for user-defined initial terms and to use these to improve the results of semi-supervised topic modelling.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...), investigators often expect specific topics relevant to the case to be discussed in these messages. Accordingly, they are interested in finding clues or evidence for the suspected topics in the immense volume of messages. Semi-supervised topic modelling can be promising. This incorporates case-specific knowledge to extract topics more aligned with the investigators' expectations. The aim is to compare known probabilistic algorithms of semi-supervised topic modelling (e.g. keyATM, Seeded LDA) concerning the extraction of case-relevant topics from forensic communication data.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...) on mobile devices, investigators are often interested in whether specific topics relevant to the case were discussed. Semi-supervised algorithms for topic modelling can be promising for this purpose. However, the linguistic structure of chat messages and the fact that they contain many high-frequency filler words present a challenge for most algorithms. One exception is the 'Guided Topic Noise Model', developed specifically for colloquial chats. This algorithm will, therefore, be used and evaluated to detect case-relevant topics in forensic chat messages.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...), investigators are often interested in whether specific topics relevant to the case were discussed in these messages. There is potential in semi-supervised topic modelling approaches to find indications of these suspected topics. Most semi-supervised algorithms are extensions of probabilistic latent Dirichlet allocation. A promising alternative is approaches that consider words' semantic similarity. This can be realised using word embeddings (e.g. word2vec, fastText). Therefore, the aim is to compare different semi-supervised approaches based on word embeddings to identify case-relevant topics in forensic chat messages.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
When analysing messages from messenger services (e.g. WhatsApp, Signal, ...) on mobile devices, investigators often expect specific topics relevant to the case to be discussed in these messages. Semi-supervised topic modelling algorithms can be used to support them in their search for clues or evidence on these topics in the immense amount of messages. Most of these algorithms are extensions of probabilistic Latent Dirichlet Allocation. A promising alternative is the algorithm 'Anchored Correlation Explanation' (Anchored CorEx), based on information theory. The aim is to apply and evaluate this algorithm for detecting case-relevant topics in forensic chat messages.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
The immense volume of messages from messenger services (e.g. WhatsApp, Signal, ...) that need to be analysed for forensic investigations presents investigators with an increasing challenge. The automatic extraction of topics can help quickly gain an overview of the content discussed in the chats. However, one problem with most topic modelling algorithms is the interpretation of the resulting topics. Usually, these are presented as a ranking of words or phrases, but without further context, they have little meaning. The idea is, therefore, to represent the topics by their most characteristic messages instead. Various similarity measures (e.g., scalar product and similarity based on word embeddings) are planned to be compared to identify the most similar messages for a topic.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
The immense volume of messages from messenger services that need to be analysed as part of forensic investigations presents investigators with an increasing challenge. The automatic extraction of topics can be promising for gaining an overview of the content discussed in the chats. However, the interpretation of these topics often proves to be problematic. Most approaches to topic modelling present them as a ranking of words or phrases that have little meaning without further context. In contrast, describing each topic by its most characteristic messages would be more helpful. Care must be taken to ensure sufficient transparency concerning subsequent usability in court when selecting the methodology for extracting these messages. The aim is, therefore, to use an Explainable Artificial Intelligence (XAI) method, the Learning Vector Quantisation algorithm, to identify the messages that are characteristic of a topic.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
The immense amount of communication data that needs to be analysed as part of investigations is proving to be an increasing challenge. The automatic extraction of topics can be helpful to gain an overview of the content discussed in the chats. However, there is often the problem of interpreting these topics, usually represented as a ranking of words or phrases with little meaning without further context. A summary describing the content of the topic would be more helpful. Therefore, the aim is to use a large language model (e.g. ChatGPT) to generate a summary for each topic automatically.
Internship and Bachelor's thesis/ Master's thesis. Contact:
Michael Spranger
More about this thesis
Messages from messenger services (e.g. WhatsApp, Signal, ...) stored on mobile devices have become an essential source of evidence in forensic investigations. However, the immense volume of messages that investigators are regularly confronted with is proving to be an increasing challenge. To gain a quick overview, it would be advantageous to summarise the messages automatically. Accordingly, the aim is to illustrate the current state of the art of methods for automatic text summarisation through a systematic literature review.
Bachelor's thesis. Contact:
Michael Spranger
Internship and Bachelor's thesis. Contact: Michael Spranger
Internship and Bachelor's thesis. Contact: Lukas Jaeckel
More about this job offer:
Staatsanwaltschaft Halle (Saale) (Zentralstelle zur Bekämpfung der Hasskriminalität im Internet). Contact:
Nadine Friedewald
More about this job offer:
Justiz des Landes Sachsen-Anhalt. Contact:
Nadine Friedewald
More about this job offer:
GICON® - Großmann Ingenieur Consult GmbH. Contact:
Nadine Friedewald