Online services (e.g. search engines, location-based services, video on demand, recommendation engines) require a massive amount of personal data harvested from their users. These data, such as location coordinates or search request history, can reveal highly sensitive user information such as political, religious, or sexual preferences. I will describe the risks stemming from the use of online services with respect to personal data and I will introduce a prospective approach for personal data protection based on decentralization and the use of hardened CPUs (e.g. Intel SGX). I will focus on two specific applications: Web requests and VOD systems.