Machine Learning (ML) pipelines rely heavily on high-quality data processing, including normalization, anomaly correction, and formatting, to ensure accurate and reliable model performance. However, many datasets used in ML applications are confidential and private, posing significant challenges to traditional data preparation methods. This internship addresses the critical need for innovative solutions that enable secure and efficient data preparation without compromising data privacy.

The primary goal of this internship is to design and implement a solution for oblivious data preparation. This approach would allow the modification and cleaning of sensitive datasets without requiring direct access to the raw data. By leveraging technologies such as Trusted Execution Environments (TEE) and trusted hardware, the intern will explore methods to ensure that data remains secure throughout the preprocessing pipeline while still meeting the stringent requirements of ML workflows.

This internship offers an exciting opportunity to work at the intersection of data privacy, security, and machine learning. The selected candidate will gain hands-on experience with cutting-edge technologies, contribute to solving real-world privacy challenges, and help pave the way for secure ML applications in industries where data confidentiality is paramount. Ideally, this project will lead to a publication at a top academic venue.

• Student enrolled in a PhD of Computer Science/Engineering

• Strong programming skills in Python and ML frameworks (Py Torch / Tensor Flow)

• Experience with implementation of ML pipelines is highly desirable

• Familiarity with TEE, cryptographic protocols and/or secure systems is a plus

• A strong publication record is a big plus

• Language skill: English

• You will get familiar with the problem by studying existing state-of-the-art regarding data preparation in ML workflows and confidential computation techniques.

• You will develop a prototype solution for confidential data cleaning, including defining the threat model and system assumptions relevant to this problem.

• You will be involved in writing an academic paper describing the solution.

Location: Stuttgart (Germany)