INFO 188 Lecture Notes - Lecture 12: Differential Privacy, Netflix Prize, Skewness
27 Dec 2018
School
Department
Course
Professor
Document Summary
Some concepts of privacy we"ve seen so far. There"s no one way to define privacy that will work in every case. Privacy arises in specific situations, in specific contexts and networks of relationships. By obfuscation, we mean producing misleading, false, or ambiguous data to make data gathering less reliable and therefore less valuable. (finn & nissenbaum) By obfuscation, we mean producing misleading, false, or ambiguous data to make data gathering less reliable and therefore less valuable. Techniques try to limit (or eliminate) the number of certain types of identifiers. Obscure too many things, and we lose many practical aspects of the dataset. Obscure too little, and we may not hide enough qualities. It is often hard to reason about all the things an adversary can do. Framework to identify potential threats from an adversaries point of view. Reason about probable attacks, vulnerabilities, and possible solutions. We will see example attacks as this lecture progresses.
