Masterarbeit: Aktuelle Backdoor Detektionsverfahren angreifen

The Fraunhofer Society is seeking a Master's thesis student to work on attacking current backdoor detection methods in neural networks. The student will develop new machine learning approaches to increase the security of large language models and present the results in a master's thesis.

Requirements

• Knowledge in the field of machine learning, including training, inference, and optimization of transformer architectures
• Good Python knowledge, in particular with PyTorch
• Scientific interest and interest in current research projects

Benefits

• Independent time allocation
• Insights into the intersection of academic research and industrial application