Privacy-preserving recommendations for Online Social Networks using Trusted Execution Environments

• Guilmour Rossi
• Luiz Gomes-Jr
• Marcelo Rosa
• Keiko Fonseca

Resumo

Online Social Networks (OSN) have changed how individuals interact with each other and with organizations, offering means of communication, publication and consumption of information. As OSNs have become a substantial part of users' online activities, OSN providers have understood the value of the data being generated and exploited it to maximize profits. Recently, malicious agents have invested in the manipulation of OSN data to attain commercial advantages or influence public opinion with dangerous consequences. This paper describes our ongoing efforts towards the use of Trusted Execution Environments (TEE), more specifically Intel's SGX, for the management of recommendation engines for OSNs. Our solution focuses on protection of user data and prevention of misuse without compromising OSNs' functionality nor OSNs' revenue from advertisements. We describe the architecture of our system and report performance results that can be used to guide the selection of recommendation algorithms for execution under SGX.