Frequent Itemset Mining Algorithm Based on Differential Privacy in Vertical Structure

Shigong Long ； Hongqin Lu ； Tingting Chen ； Nannan Zhou ； Hai Liu

《International Journal of Network Security》 24卷1期 (2022/01) Pp. 75-82

https://doi.org/10.6633/IJNS.202201_24(1).09

摘要

Frequent itemset mining is an important aspect of data mining. Direct publication of frequent item-set mining results may cause serious personal privacy leakage, so this article proposes a frequent item-set mining algorithm that satisfies differential privacy. This article proposes a weight truncation algorithm for transaction truncation to reduce the global sensitivity of differential privacy. The truncated transaction retains the frequent item information of the original transaction as much as possible. This article uses average support and maximum support estimation strategies to reduce errors due to transaction truncation and pruning operations. The experimental results are compared and analyzed, and the proposed algorithm satisfies the differential privacy protection. At the same time, the frequent itemset mined has good utility. The proposed algorithm has higher availability than the existing algorithm TT (Transaction Truncation) and algorithm PB (PrivBasis).

關鍵字

Differential Privacy ； Frequent Itemset ； Support Estimation Strategies ； Transaction Truncation