Differential privacy is a relatively new area of computer science research that has only recently been explored. Currently much of the research on privacy mechanisms focuses on static databases (i.e. relational databases). Some papers have discussed stream databases, but relatively few discuss the possibility that privacy-preserving mechanisms other than the counting mechanism could be applied on databases. While a privacy-preserving counting mechanism is a very useful aggregate in analysis of sensitive data, the proposed algorithms severely limit the scope of analysis.

For a special topics class on data privacy taught by my CS adviser Professor Katrina Ligett, we aimed to widen this scope by proposing less restrictive mechanisms, as well as aggregates other than count, such as sum, average, and variance. More background and motivation for the problem can be found in our initial project proposal here; statements and detailed proofs of our results can be found in our final report here.