How about secure data outsourcing? We have cloud databases, but they are fundamentally insecure. They don't let clients verify the integrity and completeness o

How about secure data outsourcing? We have cloud databases, but they are fundamentally insecure. They don't let clients verify the integrity and completeness of query results (correctness), they don't support the data to be queried and updated without decrypting it on the server (confidentiality) and even if they would encryption is not enough because of the leakage-abuse attacks (access privacy). https://www.vldb.org/conf/2007/papers/tutorials/p1431-sion.pdf https://www.vldb.org/conf/2007/papers/tutorials/p1431-sion.p... Currently this is an open problem. The state-of-the-art is structured encryption with forward and backward privacy and leakage perturbation for the encrypted index (but not the encrypted data). A persistent adversary with (known-data, known-query) or without (statistical inference) auxiliary information can recover data with the search pattern (frequency and co-occurrence of queried keywords), the access pattern (queried keyword to files), or only the response length (queried keyword to number of files). Oblivious RAM (ORAM) can make storage accesses indistinguishable, but it doesn't hide the response length and can't be used trivially for searchable encryption because the ORAM accesses are not indistinguishable. More importantly ORAM has a non-trivial logarithmic overhead, even with differential privacy. A constant loss in efficiency can be only had with a logarithmic loss in privacy and the other way around. Secret-sharing and trusted hardware-based techniques are similarly either inefficient or insecure. They support non-interactivity (queries and updates in one communication round), but make correctness, confidentiality and access privacy harder to achieve. I'm working on a privacy-preserving cloud database that uses volume-hiding encrypted multi-maps and query smoothing techniques (fake queries, prefetching or buffering) for a practical double logarithmic overhead on range queries and meaningful privacy, as long as the client has persistent storage, which can be used offline too. Basically client storage let's you increase privacy without compromising efficiency. There are always new things about secure data outsourcing that are not or barely covered by existing research, like maliciously secure access control or how to protect encrypted and non-encrypted indexes and data when accessed together. Or horizontal scalability (replication) without privacy loss. Or how to provide post-compromise security like history independence, secure deletion without privacy loss.