| Summary: | 碩士 === 國防大學理工學院 === 資訊科學碩士班 === 98 === Dynamic Database is widely used in various fields, such as supermarkets, banking, airline, military procurement, and security operating center (SOC) systems. Association rule analysis is a data mining technique that is easily used to find sensitive association rules in a huge database. How to prevent the risk of such information leakage is an important issue for an information system with a dynamic database.
Recent research for hiding association rule almost focused on static database environment. With scanning database once, it is able to hide association rule with minimum side effect. However, the real transactional databases are dynamic. Although a dynamic database could be simulated by a static one with a batch process, the user could not obtain real-time feedback information.
In a dynamic database environment, Support and Confidence of a sensitive association rule will change if new transactions arrived. A sensitive association rule should be completely hidden in real time. This can be done by keeping its Support or Confidence lower than their respective thresholds. The goal of this study is to hide the sensitive information with minimum side effects while guaranteeing real-time responses from information systems.
In this study, we completely analyze all types of schemes and propose four on-line algorithms for association rule hiding, 1) ANT (Add New Transaction), 2) DRI (Delete Right hand side Item), 3) DLI (Delete Left hand side Item), and 4) CLI (Complement Left hand side Item). Finally, we develop an adaptive algorithm, ATM (Adaptive Transaction Modification). The experiment results show that ATM outperforms other algorithms in terms of overall side effects.
|
|---|
