NotesFAQContact Us
Search Tips
ERIC Number: ED530719
Record Type: Non-Journal
Publication Date: 2011
Pages: 165
Abstractor: As Provided
Reference Count: 0
ISBN: ISBN-978-1-1247-0992-5
Target-Based Maintenance of Privacy Preserving Association Rules
Ahluwalia, Madhu V.
ProQuest LLC, Ph.D. Dissertation, University of Maryland, Baltimore County
In the context of association rule mining, the state-of-the-art in privacy preserving data mining provides solutions for categorical and Boolean association rules but not for quantitative association rules. This research fills this gap by describing a method based on discrete wavelet transform (DWT) to protect input data privacy while preserving association rules in static and dynamic environments. The key idea is to sort the data on each attribute, perform one-level Haar transform on each sorted attribute, retain approximation coefficients and copy them to two consecutive rows. In the static case, the strategy is simple since changes to the collection of data are ignored. In the case where data is dynamic, updates must be captured incrementally to save computational and I/O resources. Preserving association rules and the privacy of data in a dynamic environment is a three-step process. In step one, an approach to synchronize a source location with a target location is proposed. Known as Target-Based Database Synchronization (TBDS), this approach is aimed at solving a variation of the synchronization problem. The approach relies on comparing hash values for corresponding partitions of the source and the target databases so that if these values differ, the tuples in the target partition can be dropped and reloaded from the source. A table of partition information guides this process. In step two, the target-based database synchronization approach is used in conjunction with the discrete wavelet transform approach to incrementally update transformed data. Finally in step three, deleted and inserted transformed deltas retrieved from the data owner's database transaction logs are passed to the data miner who combines them with his previous set of frequent itemsets to efficiently compute the new set of frequent itemsets and rules. Experimental results demonstrate that compared to existing approaches such as kd-tree and random projection, the proposed approach renders higher accuracy and privacy for small to moderate size of database updates when updates lie within attribute domains, preserves 90%-100% of the original rules regardless of whether the size of database updates is small or moderate and executes efficiently when the percentage of database changes are small. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page:]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site:
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A