Data Accountability

Several data-protection regulations, such as the right to be forgotten, presuppose reasoning about data accountability. For example, when a user requests her data to be deleted from a system, one first needs to determine which data items have been computed based on this user’s data, and then delete only those items. So, one needs to determine for which items the user’s input data is accountable for. This determination becomes challenging when input data of several users is correlated. Ideally, an item should be deleted only because it was actually computed based on certain user’s input---not because it is correlated with certain user’s input. We formulate a novel theory for defining data accountability.