Go to JKU Homepage
The core question of this work is whether (and if so, how) the learning process of an ML model can be made traceable in an efficient way so that the results of the training can be reproduced reliably. Which algorithms are suitable for this purpose? What impact would this have on the security of AI?
Teaching.
Current Courses
This semester, we offer the following Lectures (VO, VL), Exercises (UE) and Combined Courses (KV).
| {{ labelInLang('cid') }} | {{ labelInLang('title') }} | {{ labelInLang('registration') }} | {{ labelInLang('type') }} | {{ labelInLang('hours') }} | {{ labelInLang('teachers') }} | {{ labelInLang('rhythm') }} |
|---|---|---|---|---|---|---|
| {{ item._id }} ({{ item.term }}) |
{{ item.title }}: {{ item.subtitle }}
{{ labelInLang('moreinfo') }} {{ labelInLang('expand') }} {{ labelInLang('collapse') }} |
{{ labelInLang('register') }} | {{ item.type }} | {{ item['hours-per-week'] }} | {{ teacher.firstname }} {{ teacher.lastname }} {{ item.teachers.teacher.firstname }} {{ item.teachers.teacher.lastname }} | {{ item.rhythm }} |
| {{ item._id }} ({{ item.term }}) | |
| {{ labelInLang('title') }} |
{{ item.title }}: {{ item.subtitle }}
{{ labelInLang('moreinfo') }} {{ labelInLang('expand') }} {{ labelInLang('collapse') }} |
| {{ labelInLang('registration') }} | {{ labelInLang('register') }} |
| {{ labelInLang('type') }} | {{ item.type }} |
| {{ labelInLang('hours') }} | {{ item['hours-per-week'] }} |
| {{ labelInLang('teachers') }} | {{ teacher.firstname }} {{ teacher.lastname }} {{ item.teachers.teacher.firstname }} {{ item.teachers.teacher.lastname }} |
| {{ labelInLang('rhythm') }} | {{ item.rhythm }} |
Internships & Scientific Theses (Seminar, Bachelor’s, Master’s)
The following topics are currently available. Details on each topic can be found below; further information can be provided in a personal meeting. If you are interested, please contact us by email (links provided below).
Topics on Security of AI
The goal of this work is to survey attacks on ML models during the training phase. How can the deliberate introduction of noise in the training data be counteracted? Can algorithms be influenced by their (other) configuration options in addition to the input data?
The core of the work is the question whether an LLM can be "manipulated" with artificially generated messages to such an extent that it produces unintended or deliberately induced results or would even pass on falsely predicted messages (fake news)? In the context of the work, an experimental LLM is to be set up, and manipulated with a second one (adversarial LLM).
Topics from Cryptography
Large networks can be modeled as random graphs, but their statistical distribution is a question answered inconsistently in the literature. The goal of the Master's thesis is to statistically test for different distribution models for random graphs.
Deep knowledge in statistical test theory is not required; the statistical tests can be performed with standard tools.
The goal of the master thesis is to survey access rights to different files in shared storage systems like clouds. Which person has access to which files? How can these access rights be efficiently determined and kept up to date? The goal of the thesis is an overview of procedures and tools that allow such surveys.
The aim of the work is an introduction and an overview of current developments in the mentioned field of work. If enough sources are used, the thesis, if applicable, can be published as a survey paper in scientific journals.
The aim of the work is an introduction and an overview of current developments in the mentioned field of work. If enough sources are used, the thesis, if applicable, can be published as a survey paper in scientific journals.
The aim of the work is an introduction and an overview of current developments in the mentioned field of work. If enough sources are used, the thesis, if applicable, can be published as a survey paper in scientific journals.
The aim of the work is an introduction and an overview of current developments in the mentioned field of work. If enough sources are used, the thesis, if applicable, can be published as a survey paper in scientific journals.
Other topics and topics of your own suggestion
The topics listed below are to be understood as “contexts”, within which concrete ideas and topics can be discussed. Your own suggestions for thesis topics are of course also more than welcome. We look forward to a detailed discussion if you are interested.
All these topics lend themselves to define internships, project practicals, seminars, bachelor’s or master’s theses:
- Extensions to our crypto programming language Sunset/FFapl, opens an external URL in a new window. This includes both, language extensions and additions/changes to the graphical user interface or the development environment, the interpreter or functions of the language.
- Applied Cryptology
- Functional encryption
- Private/Secure function evaluation
- Quantum Cryptography
- Functional digital signatures
- Special-purpose cryptography
- …your own suggestions/interest in Applied Cryptology…
- Risk Management and Applied Security
- Game-theory for defense against advanced persistent threats (prototyping of models, implementation of simulation testbeds, …)
- Graphical and other methods of risk elicitation and threat modelling
- Application of artificial intelligence
- Security of artificial intelligence
- Simulation models for incident prediction and propagation
- Digital twins for security
- Cyberwarfare
- …your own suggestions/interest in Risk Management and Applied Security…
Mandatory coursework during the practical/internship/thesis writing phase
Depending on the current level of your studies, you must enrol in one of the following courses:
Type of work | Course to enrol |
Seminar | 510.503 Seminar in Networks and Security Cryptography and Security Infrastructures |
Bachelor’s Thesis | 353.055 Projektpraktikum |
Project (Bachelor’s) | 353.055 Projektpraktikum |
Master’s Thesis | 353.095 Master's Thesis Seminar (summer term) and/or 353.094 Master's Thesis Seminar (winter term) |
Internship or Project (for Master’s) | 353.093 Project in Networks and Security |
Presentation dates and types
In all of the above courses, you will be asked to present the status or progress of your work in an oral presentation. Depending on how far you are, you need to give (just) one of three presentations:
- a "starter presentation", showing what the project is about and what the first steps are going to be (15min duration), or
- a "current status presentation " showing what has happened so far and what the next steps are (20min duration), or
- a "finish presentation ", presenting the results and wrapping up the project (30min duration).
All three presentations should include at most 5min for questions and answers (leaving the respective times for the presentation as at least 10min for the start, 15min for the intermediate, and 25min for the finish talks).
Dates and times are fixed individually with your supervisor.
Tips for writing your thesis
We have put together some best practices, do’s and don’ts for writing the thesis, searching for literature and other miscellaneous matters. See the document provided, and use it as an aid and guideline for the writing.
Further helpful information to write and submit your Bachelor's/Master's thesis, or dissertation is available here.
