Tunable Stream Graph Embeddings at Scale
- Tunable Stream Graph Embeddings at Scale” Serafeim Papadias (TU Berlin), supervised by Volker Markl (TU Berlin / DFKI)
- Ph.D Workshop at the 46th International Conference on Very Large Data Bases (VLDB 2020)
- Download the paper: http://ceur-ws.org/Vol-2652/paper10.pdf
Although the cloud is today a de-facto standard for scalable data processing, there are still many applications that cannot make use of the cloud due to data or computation privacy. Sensitive data, such as in the health do-main; and computations, such as core-business AI pipelines, grew into valuable assets that made secure data processing a hot topic in industry and academia. On one hand, the existing data processing systems prioritize performance and, to a certain level, trade users’ privacy. On the other hand, privacy-preserving data processing systems sacrifice performance. In this PhD thesis, we envision a fully secure general-purpose data processing system for the cloud. Over-all, we aim at devising: (i) algorithms that are adequate to work with very limited memory, such as the one exposed by trusted execution environments; (ii) scalable state management techniques; (iii) oblivious data-access algorithms; and (iv) privacy-preserving query optimizations techniques to speed up query execution.