Towards an Era of Trust in Personal Data Management
Résumé
Managing personal data with strong privacy guarantees has become an important topic in an age where your glasses record and share everything you see, your wallet records and shares your financial transactions, and your set-top box records and shares your energy consumption, while several recent affairs have unveiled the severe consequences of the loss of privacy. In this context, more and more alternatives are proposed based on user centric and decentralized solutions, capitalizing on the use of trusted personal devices controlling the data at the edges of the Internet. Decentralized solutions are promising because they do not exhibit the intrinsic limitations of classical centralized solutions, e.g., sudden changes in privacy policies of companies holding the data, data exposures by negligence or because it is regulated by too weak policies, exposure to sophisticated attacks whose benefit/cost ratio is high for centralized databases. Hence, such solutions appear as a sea change for personal data management, where the control over personal data is pushed to the edges of the Internet, within sensors acquiring the data and in a variety of user devices endowed with a form of trust, e.g., tamper-resistant secure hardware-based devices.
This tutorial reviews several existing solutions going in this direction, presents a functional architecture encompassing these alternatives, and exposes the underlying techniques and open issues dealing with user centric and decentralized data management platforms. In a first part, we review the recent initiatives pursuing the objective of reestablishing user control over their data by decentralizing this control in personal secure or trusted devices. We discuss an abstract distributed architecture focusing on secure storing, managing and sharing of personal data, i.e., the asymmetric architecture, and indicate the main challenges inherent to decentralized data management. In a second part, we explore data management techniques exercised within a trusted device at the client side. We review the main attempts proposed in the literature and concentrate on those addressing the specific context of microcontrollers equipping sensors and mobile phones (SIM cards). In a third part, we investigate the problem of performing global processing without any compromise on data privacy. We present the difficulties to overcome to execute privacy preserving computations on populations of personal devices, and illustrate it by focusing on Group By SQL queries and Privacy Preserving Data Publishing. In a fourth part, we conclude the tutorial by presenting existing and future instances of decentralized privacy preserving data management architectures. We mainly focus on attempts and proposals targeting social-medical, smart houses, and rural areas contexts.