Training of Artificial Intelligence (AI) algorithms requires collecting large volumes of training and reference data to feed analytics and predictive models. However, there is often lack of the training data for machine learning techniques. The scarcity is caused, generally speaking, by reluctance to share data by third parties for confidentiality, privacy and other reasons (i.e. economic value of data).

Resolving the problem of systematic access to training and reference data has been therefore recognised as a stepping stone to realise the true potential of AI for EO in the Big Data Platform architectures. Privacy Preserving Machine Learning (PPML) is an emerging field in data science that strives to address this challenge with a variety of new approaches to machine learning currently available such as Federated Learning (FL), Secure Multi-Party Computation (MPC), Homomorphic Encryption (HE), Differential Privacy (DP) and Zero-Knowledge Proofs (ZKP), among others.

Therefore as large repositories of different economic, financial and social data (i.e. household and census surveys), including location data collected via mobile phones and mobile apps are being currently made available, the new ESA ITT opens opportunities for EO community to conduct research and development in terms of how to apply these in EO analytics in a privacy-preserving way.

Learn more about this Invitation To Tender on the EMITS website.