When “capacity” changes with set size: Ensemble representations support the detection of across-category changes in visual working memory

Schurgin, Mark W.; Brady, Timothy F.

doi:10.1167/19.5.3

We are storing and querying datasets with the private information of individuals at an unprecedented scale in settings ranging from IoT devices in smart homes to mining enormous collections of click trails for targeted advertising. Here, the privacy of the people described in these datasets is usually addressed as an afterthought, engineered on top of a DBMS optimized for performance. At best, these systems support security or managing access to sensitive data. This status quo has brought us a plethora of data breaches in the news. In response, governments are stepping in to enact privacy regulations such as the EU’s GDPR. We posit that there is an urgent need for trustworthy database system that offer end-to-end privacy guarantees for their records with user interfaces that closely resemble that of a relational database. As we shall see, these guarantees inform everything in the database’s design from how we store data to what query results we make available to untrusted clients. In this position paper we first define trustworthy database systems and put their research challenges in the context of relevant tools and techniques from the security community. We then use this backdrop to walk through the “life of a query” in a trustworthy database system. We start with the query parsing and follow the query’s path as the system plans, optimizes, and executes it. We highlight how we will need to rethink each step to make it efficient, robust, and usable for database clients.

More Like this