Multi-level high utility-itemset hiding.

Area of Science:

• Data Mining
• Information Security
• Database Management

Background:

• Data privacy is a critical concern, especially with sensitive information in transaction databases.
• High-utility itemset mining (HUIM) identifies valuable patterns but can expose sensitive data.
• Existing privacy-preserving utility mining (PPUM) methods often overlook data generalization, leaving higher abstraction levels vulnerable.

Purpose of the Study:

• To develop novel PPUM algorithms that protect sensitive high-utility itemsets across all abstraction levels.
• To address the limitations of current PPUM techniques that focus only on specialized item levels.
• To enhance data security for organizations sharing transaction data.

Main Methods:

• Introduction of two new PPUM algorithms: MLHProtector and FMLHProtector.
• Designing algorithms to operate effectively at multiple levels of data abstraction.
• Empirical evaluation of algorithm performance in protecting sensitive itemsets.

Main Results:

• MLHProtector and FMLHProtector successfully protect sensitive high-utility itemsets from unauthorized discovery.
• The proposed algorithms ensure data confidentiality across various abstraction levels.
• Experimental results validate the effectiveness of the developed PPUM techniques.

Conclusions:

• The developed MLHProtector and FMLHProtector algorithms offer robust privacy protection for high-utility itemsets in transaction databases.
• Addressing all abstraction levels is crucial for comprehensive privacy preservation in utility mining.
• These algorithms provide a significant advancement in securing sensitive business data against data mining threats.