Biometric recognition of individuals based on the observation of behavioural and biological characteristics, such as face, iris, or fingers is widespread in access control applications. Such methods can be used for verification or identification applications and constitute a comfortable alternative to knowledge based or token based methods.

In the last decade, the availability of larger amounts of data and higher computational resources has led to a revolution in machine learning, leading to the so called deep learning architectures. These algorithms can provide higher classification accuracy or generate very realistic synthetic images or audios.

Current advances in AI and machine learning allow us to automate certain tasks or enhance the obtained results. However, processing personal data without appropriate safety measures can also lead to undesirable disclosure of sensitive information. Indeed, machine learning models often contain precise information about individual data points that were used to train them. We thus have to protect the privacy of the individuals, trying at the same time to maximise the utility of the machine learning model.