How Private is Your Data Analysis?
Sara Krehbiel
Assistant Professor, Department of Mathematics and Computer Science Santa Clara University
Stevenson Hall 1300
11:00 AM
- 11:50 AM
Our personal data is collected for many purposes. How can we impose safeguards on data analysis to make sure it doesn't inadvertently leak individual data? Differential privacy is a mathematical guarantee that even if data analysis reveals rich information about a population, there is little it can reveal about an individual, even if an adversarial party knows almost everything about the dataset. But how much can a more realistic adversary learn? We explore how this depends both on what the adversary already knows and on the algorithms used to achieve differential privacy in the first place.