In the last decade the amount of sensitive data that is stored in the cloud has skyrocketed. On the one hand, off-site storage provides undeniable conveniences, as for example location independent data access and the certainty, that a hardware malfunction does not imply data loss. On the other hand, cloud storage raises some negative issues as well. The most prevalent of which is confidentiality against the storage host as well as third parties. Methods to ascertain confidentiality boil down to encryption and secret sharing. Both have their respective dis- and advantages, especially regarding computation on the stored data. Gentry famously presented his FHE scheme, enabling computation on encrypted data, while for computation on secret shared data, there are multiparty computation protocols, that maintain confidentiality.
In our work we investigate how to (more) efficiently execute computations on secret shared data without reconstructing it. The computation will be performed by two external computing units in an MPC setting, that obtain shares of the secret from a minimal set of shareholders. Thereby we circumvent the secret's reconstruction. Furthermore we achieve maximum confidentiality, that is the computation units only learn the executed computation but not the secret's identifier and the shareholders only learn the identifier but not the computation.
Philipp Muth is a PhD student of Stefan Katzenbeisser in the Security Engineering group. His research focuses on different aspects of long term security, on which he works within the CROSSING Project S6.
We will have lunch (pizza) together after the research seminar talk. Please register here: https://doodle.com/poll/ak29ykuuercv976p