Privacy-preserving data publishing: K-anonymity and l-Diversity

This lecture covers the concepts of K-anonymity, database sanitization, and l-Diversity in the context of privacy-preserving data publishing. It discusses the challenges of k-anonymity, the weaknesses of this approach, and introduces l-Diversity as a solution. The presentation delves into the limitations of f-diversity and l-diversity, highlighting the importance of considering the overall distribution and semantics of sensitive values. Real-life examples, such as the Netflix dataset release, illustrate the failures of naive de-identification methods and the risks associated with sparse high-dimensional data. The lecture concludes by examining the case of Airbnb's data privacy efforts and the potential vulnerabilities in their approach.