Nowadays there is a great interest in developing techniques assessing the user privacy risk due to the requirements of the GDPR.
Traditional frameworks conduct an evaluation of the privacy risk by simulating a series of possible privacy attacks.
The main drawback is their efficiency. We propose a user-centric privacy risk assessment framework that employs machine learning models for predicting the users’ privacy risk. Our interest is on time series data. Our framework also involves a privacy risk explanations component. Its main task is to provide an explanation to the end-user, describing the reasons why his data put him at risk. The proposed approach solves the problem of efficiency for the privacy assessment methodology and gives the opportunity to assign the privacy risk to a user without looking at the data of other users.
Experimental results highlight good predictive performances and are promising in terms of explainability for increasing the user awareness.