Re-identifying people from anonymous histories of their activities

Abstract

Privacy problems are major obstacles to collecting and using big data because, in many cases, big data reflects a person’s history of activities, such as moving around a city, buying goods, surfing the Web, and posting content on social media. Although anonymization is an effective technical measure for alleviating privacy concerns, we must be aware of two problems that could infringe privacy: re-identifying the people represented by the data despite anonymization and profiling people from the data. In this paper, we first survey reidentification techniques developed for various areas, clarify the relationship between re-identification and profiling, and mathematically model the re-identification problem. We then present methods for re-identifying social media accounts and location histories and present the results of evaluations demonstrating their effectiveness.