In a data-driven world, the integration of artificial intelligence (AI) into business processes and products is becoming increasingly important for software companies such as DATEV eG. However, this poses considerable challenges for information security and data protection, as long as AI systems process confidential data. In this presentation, we briefly introduce concepts and applications of Privacy-Preserving Machine Learning (PPML) in AI systems. PPML makes it possible to train models and make predictions without revealing sensitive information from the model.

The focus of the presentation is on explaining the basic principles of differential privacy using an example and its implementation. IT security-relevant topics, such as security features of and attacks on PPML, wil ...