Technical Perspective: Graph Theory for Data Privacy: A New Approach for Complex Data Flows

Abstract

Nearly all of the world's population now uses online services that request personal information, covering almost every aspect of our lives. The abundance of personal data in digital form has brought incredible benefits to end users, enabling them to access personalized and advanced services based on the analysis of the data collected. This capability has dramatically improved the user experience in various application domains, ranging from healthcare to e-commerce, finance, logistics, and entertainment, to name a few. Numerous technological advancements in the field of big data have enabled this massive processing of personal data, and recent advances in AI data processing capabilities will expand the ways in which service providers will use personal data in the coming years. Machine learning algorithms, powered by AI, will be used to make increasingly accurate predictions about user behavior by uncovering hidden correlations within massive data sets. There is therefore a tension between the desire to fully exploit personal data in such ecosystems and the need to provide strong privacy and transparency guarantees to the individuals whose data is being exploited. Privacy protection is further complicated because data processing is typically not performed in isolation but through pipelines of different services, with each step making inferences about the personal data consumed by the services in subsequent steps.