关键词: 隐私保护, l-多样性, 敏感度, 聚类, 个性化, 相似性攻击

Abstract: Currently,the majority of anonymity model for privacy preservation neither meets the need of personalized preservation oriented to sensitive attribute values,nor considers the distribution of sensitive attribute values,therefore they are vulnerable to similarity attack.This paper proposes a personalized (α,l)-anonymity model based on sensitivity.This model provides sensitivity for different sensitive attribute values,and defines the concept of equ-sensitivity group,implements the personalized needs of privacy anonymity by setting the frequency constraints for different equ-sensitivity group in every equivalence class.This model also can defense similarity attack by limiting the total frequency of sensitive attribute values for same sensitivity in every equivalence class,and control the distribution of sensitive attribute values.This paper proposes a personalized (α,l)-anonymity algorithm based on clustering to achieve the purpose of anonymity.Experimental results show that the proposed algorithm provides better privacy preservation than other l-diversity anonymity models with the similar information loss and time cost.

Key words: privacy preservation, l-diversity, sensitivity, clustering, personalized, similarity attack