Organizer: Giorgia Azzurra Marson / Nina Bindel
As fully sequenced human DNA becomes an inexpensive commodity, it will not come as a surprise in the near future to see thriving business models, that will use sequenced DNA as their driving force. In such case it will be desirable, that the fully sequenced human genome be outsourced encrypted (due to the data's highly private nature) to a remote biobank, so that it will be processed by an authorized party (here called an investigator) on demand and in large cohorts
A closer look at the problem however makes apparent that encryption alone is not enough to guarantee the owner's privacy: The biobank (or an investigator) can deduce enough information about previous processing on the data by merely studying the access patterns (i.e. which parts of the data have been previously accessed). In order then to preserve the data owner's privacy, a cryptographic primitive proposed by Goldreich and Ostrovsky, known as Oblivious RAM (ORAM) seems to be a promising candidate as a building block for solutions to the above described problem.
Inspired by the ORAM primitive, in this talk we will revise the state-of-the-art ORAM constructions proposed by Shi et al. and explore their ability to form the building blocks for solutions, that can guarantee access pattern privacy in processing of encrypted fully sequenced human DNA. Furthermore we investigate their potential usage not only in settings where multiple clients store their encrypted DNA, but also in settings where multiple investigators are allowed to process the encrypted DNA of multiple data owners.