关键词: 频繁模式挖掘, 隐私保护, 增量更新, 部分隐藏的随机化回答算法

Abstract: Aiming at the problems that a database is scanned for multiple times and a record is compared for many times to count in most frequent pattern mining algorithms for privacy-preserving, an Incremental Bitmap-based Randomized Response with Partial Hiding (IBRRPH) algorithm was proposed. Firstly, the bitmap technique was used to represent the transaction in the database, and the "and" operator for bit was used to speed up the support degree calculating. Secondly, an incremental update model was introduced by analyzing incremental access relationship, so that the mining result before was used to the maximum limit during incremental updating. The contrast experiment of performance to the algorithm proposed by Gu et al. (GU C, ZHU B P, ZHANG J K. Improved algorithm of privacy preserving association rule mining. Journal of Nanjing University of Aeronautics & Astronautics, 2015, 47(1):119-124) was done aiming at the increment range from 1000 to 40000. The experimental results show that the efficiency of the IBRRPH algorithm is improved over 21% compared to the algorithm proposed by Gu et al.

Key words: frequent pattern mining, privacy-preserving, incremental update, Randomized Response with Partial Hiding (RRPH) algorithm