Abstract—Information sharing among the organizations promotes business growth. Recent trends in data mining techniques impose an intimidation to data sharing. A key problem here is the need to balance the privacy of the data with the genuine need for the users. To address this problem, data sanitization process modifies the original data to conceal sensitive knowledge before release. Several researchers addressed the privacy preservation of sensitive knowledge in the form of association rules by suppressing the frequent itemsets. This paper proposes an effective data sanitization algorithm which minimizes the side effects caused in the original database. A hybrid conflict ratio approach is proposed to pick the victim transactions and items to minimize the legitimate lost during sanitization. To study the effectiveness of the algorithm, experimental analysis is carried over the real and synthetic datasets. The results illustrate that the algorithm show good results compared with the other approaches.
Index Terms—Data Sanitization, Association Rule mining, Privacy Preserving Data Mining, Conflict Ratio.
R. R. Rajalaxmi is with Computer Science and Engineering Department, Kongu Engineering College, Erode, Tamil Nadu, India (phone : 04294-226580),
A. M. Natarajan is CEO of Bannari Amman Institute of Technology, Sathtyamangalam, Erode, Tamil Nadu, India.
Cite: R. R. Rajalaxmi and A. M. Natarajan, "Hybrid Conflict Ratio for Hiding Sensitive Patterns with Minimum Information Loss," International Journal of Computer Theory and Engineering vol. 1, no. 4, pp. 430-433, 2009.