To be able to handle more and more complex tasks, Deep Neural Networks need to be trained with more and more data. One strategy to obtain a large dataset in times of increasing privacy awareness is Federated Learning. Instead of requiring all data on one central place, Federated Learning outsources the training to several clients, such that each client trains its own model locally and only shares the parameters of the trained models.
The lectures will be accompanied by practical homeworks, where state-of-the-art attacks and defenses will be implemented. At the end, you will implement your own attack and defenses that have to compete against the submissions of the other participants.
In the seminar, you will summarize and evaluate existing literature for your choosen topic related to System and IoT security and report their findings in the form of a seminar paper.
Possible topics include:
This event addresses current topics from research and development with regard to security.
Analyzing and developing security solutions are complex tasks, which require knowledge from different areas of computer science. The aim of this lab is to combine skills from different areas within a project from the security sector.
Tasks from a very wide range (from algorithmics, space travel, and machine learning to software analysis, hardware development, and reverse engineering) will be presented.
The final tasks are determined individually and according to the interests/skills of the participants.
Depending on the task's scope and level, this course will be completed as a Lab (InoSys-Lab with 6CP) or as a Project-Lab (InoSys-Project with 9CP). The type will be determined individually and task-dependent. At this point, and as far as the nature of the task allows, the students will have the opportunity to participate intellectually in the design of the task.