Ecole d'ingénieur et centre de recherche en Sciences du numérique

Data protection in the era of artificial intelligence. Trends, existing solutions and recommendations for privacy-preserving technologies

Araujo, Rosa; Crespo Garcia, Alberto; Farkash, Ariel; Garnier, Antoine; Kiousi, Akrivi Vivian; Koster, Paul; Kung, Antonio; Livraga, Giovanni; Diaz Morales, Roberto; Önen, Melek; Palomares, Angel; Navia Vazquez, Angel; Metzger, Andreas

Research Report, October 2019, Timan, Tjerk & Mann, Zoltan (eds), October 2019, BDVA

One of the challenges of big data analytics is to maximize utility whilst protecting human rights and preserving meaningful human control. One of the main questions in this regard for policy- and lawmakers is to what extent they should allow for automation of (legal) protection in an increasingly digital society? This paper contributes to this debate by looking into different technical solutions developed by the projects of the Big Data Value Public-Private Partnership (BDV cPPP) that aim to protect the privacy and confidentiality whilst allowing for big data analytics. Such Privacy-Preserving Technologies are aimed at building in privacy by design from the start into the back-end and front-end of digital services. They make sure that data-related risks are mitigated both at design time and run time, and they ensure that data architectures are safe and secure. In this paper, we discuss recent trends in the development of tools and technologies that facilitate secure and trustworthy data analytics and we provide recommendations based on the insights and outcomes of the projects of the BDV cPPP and from the task forces of the Big Data Value Association (BDVA), combined with insights from recent debates and the literature. In this paper, privacy challenges are addressed that stem particularly from working with big data. Several classification schemes of such challenges are discussed. The paper continues by classifying the technological solutions as proposed by current state-of-the-art research projects. Three trends are distinguished, which are 1) putting the end user of data services back as central focus point of Privacy-Preserving Technologies, 2) the digitization and automation of privacy policies in and for big data services, and 3) developing secure ways of multi-party computation and analytics, allowing both trusted and non-trusted partners to work together with big data while simultaneously preserving privacy. The paper ends with three main recommendations: 1) the development of regulatory sandboxes, 2) the continued support for research, innovation and deployment of Privacy-Preserving Technologies, and 3) the support and contribution to the formation of technical standards for preserving privacy. The findings and recommendations of this paper in particular demonstrate the role of Privacy-Preserving Technologies as an especially important case of data technologies towards data-driven AI Privacy-Preserving Technologies constitute an essential element of the AI Innovation Ecosystem Enablers (Data for AI) as elaborated in the joint BDVA and euRobotics strategic research, innovation and deployment agenda towards a European AI partnership (AI PPP SRIDA). This paper thereby provides an elaboration of the challenges spelled out in the AI PPP SRIDA.

Document Bibtex

Titre:Data protection in the era of artificial intelligence. Trends, existing solutions and recommendations for privacy-preserving technologies
Type:Rapport
Langue:English
Ville:
Date:
Département:Sécurité numérique
Eurecom ref:6074
Bibtex: @techreport{EURECOM+6074, year = {2019}, title = {{D}ata protection in the era of artificial intelligence. {T}rends, existing solutions and recommendations for privacy-preserving technologies}, author = {{A}raujo, {R}osa and {C}respo {G}arcia, {A}lberto and {F}arkash, {A}riel and {G}arnier, {A}ntoine and {K}iousi, {A}krivi {V}ivian and {K}oster, {P}aul and {K}ung, {A}ntonio and {L}ivraga, {G}iovanni and {D}iaz {M}orales, {R}oberto and {\"{O}}nen, {M}elek and {P}alomares, {A}ngel and {N}avia {V}azquez, {A}ngel and {M}etzger, {A}ndreas}, number = {EURECOM+6074}, month = {10}, institution = {Eurecom}, url = {http://www.eurecom.fr/publication/6074},, }
Voir aussi: