Efficiently hiding sensitive itemsets with transaction deletion based on genetic algorithms.

Area of Science:

• Computer Science
• Data Science
• Information Security

Background:

• Data mining extracts valuable insights from large datasets.
• Data mining techniques pose privacy risks by potentially exposing sensitive information.
• Privacy-preserving data mining (PPDM) aims to mitigate these risks by sanitizing data.

Purpose of the Study:

• To propose a novel algorithm for effective privacy preservation in data mining.
• To address the NP-hard nature of data sanitization in PPDM.
• To minimize the side effects associated with hiding sensitive information.

Main Methods:

• Development of a compact prelarge GA-based (cpGA2DT) algorithm.
• Integration of compact genetic algorithm (cGA) and prelarge concepts.
• Design of a flexible fitness function with adjustable weights for transaction deletion.

Main Results:

• The cpGA2DT algorithm demonstrates effective hiding of sensitive itemsets.
• The algorithm balances hiding failure, missing cost, and artificial cost.
• Experimental results show competitive performance against existing methods.

Conclusions:

• The proposed cpGA2DT algorithm offers an efficient solution for privacy-preserving data mining.
• It effectively tackles the challenges of data sanitization in PPDM.
• The algorithm provides a flexible approach to managing privacy-utility trade-offs.