Research Program 4 (R4)

Privacy-preserving and secure AI

The goal of FCAI’s Research Program Privacy-preserving and secure AI is to develop methods for privacy-preserving machine learning and artificial intelligence, especially based on differential privacy. Privacy-preserving and secure AI contributes mainly to FCAI research objective Trust and ethics (objective 2). Moreover, strong privacy preservation will ease the problem of data scarcity (closely related to objective 1, data efficiency) through encouraging more data sharing.

We are very active in developing differentially private machine learning methods, especially for Bayesian machine learning used in Agile Probabilistic AI. Our work also covers cryptographic and secure multi-party computation techniques for ensuring the security and privacy of the training of AI systems and their use in prediction. We cover a number of applications from health to generic deep learning and differentially private data release.

Examples of publications:

Program poster (pdf)

Coordinating professor: Antti Honkela – antti.honkela at helsinki.fi

People

The groups of following PIs take part in the Research Program Privacy-preserving and secure AI. If you would like to join this program, please contact the coordinating professor.