中央研究院 資訊科學研究所

活動訊息

友善列印

列印可使用瀏覽器提供的(Ctrl+P)功能

TIGP (SNHCC) -- Studies on Pattern Mining and Privacy Preserving

:::

TIGP (SNHCC) -- Studies on Pattern Mining and Privacy Preserving

  • 講者蔡曉萍 教授 (國立中興大學電機工程學系)
    邀請人:TIGP SNHCC Program
  • 時間2015-09-30 (Wed.) 14:30 ~ 16:30
  • 地點資訊所新館106演講廳
摘要

As huge data are increasingly generated and accumulated, outsourcing data for storage, management, and knowledge discovery is becoming a paradigm. However, given that a lot of sensitive information and valuable knowledge are hidden in data, the outsourcing of data is vulnerable to privacy crises and leads to demands for generalization or suppressing techniques to protect data from re-identification attacks. In this talk, I would like to introduce our recent two works on pattern mining and privacy preserving.

In the first work, we explore the problem of re-identification attack on transactional database. Since frequent buying behaviors can be used to profile customers and observed easily from transactional datasets, to prevent adversaries from re-identifying a user on transactional datasets, we propose the KAMPp1 algorithm to generalize and suppress data to make the data to be published qualify the k-anonymity model. To study the effectiveness of the proposed algorithm, we conduct experiments on a synthetic and a small real dataset. The experimental results show that KAMP-p1 algorithm can satisfy k-anonymity while preserving many patterns in order to retain useful knowledge for decision making. In the second work, we study the problem of privacy preserving in trajectory pattern mining. Since most existing approaches modify locations to make the modified trajectory data confirm to the -anonymity model without considering location semantics, mining on the modified data may lead to misleading results. Also, it is unnecessary to brutally provide the same level of privacy protection to all locations. Therefore, we infer four privacy risk levels based on the risk of privacy breach and propose the Semantic Space Translation (SST) algorithm to translate locations to retain correct semantics of discovered patterns. To verify the performance of our approach, we conduct several experiments and the experimental results show that our idea is feasible and the SST can strike a good balance between privacy preserving and data utility.