基於網頁結構探勘之受限項目集隱藏

Hiding Restrictive Itemsets Based on Web Structure Mining

現代企業對於資料的分析相當重視，而龐大的資料已經難以利用人工來分析整理。因此，近年來便利用資料探勘(data mining)資料探勘技術來快速分析所累積的龐大資料，得到有用的資訊，並以提升公司的競爭力。在競爭激烈的商業環境中，企業為提高本身的生產及競爭優勢，常會與許多相關產業合作，彼此共享資料。為避免企業已知且列為機密的獨家資訊給合作的公司所探勘出來。在企業間彼此共享資料之前必需先刪除或隱藏重要的敏感或隱私資料，避免危害到公司的利益。因此，事先的隱藏處理是非常重要的。近年來，關於敏感資訊之隱藏研究，多基於刪除項目的方法。此法會導致一些非敏感的資訊同時也被隱藏，此效應稱之為邊際效應(side effect)。因此，如何有效的隱藏敏感資訊且降低邊際效應是相關研究的主軸，也是本研究的重點。本研究利用網頁結構探勘技術，將集中性網頁及權威性網頁分別對應到交易記錄及項目中，即可根據所計算出的hub及authority權重值去隱藏包含敏感資訊的受限項目集。實驗結果顯示，本方法在不同的門檻值之下有著低度的邊際效應。

Data analysis technologies are very important in modern enterprises. To analyze the collected huge data by human beings is impossible. Recently, data mining technologies have been used to analyze the hung data efficiently. In a business/competitive environment, some companies may share data for mutually benefits, and must avoid leaking their own sensitive information. Therefore, a sanitization process should be employed to hide the sensitive information before sharing the original data. In recent years, many sanitization processes focus on deleting items to hide sensitive information. However, deleting items may result in some non-sensitive information being hidden, unexpectedly. This situation is called the side effect. Therefore, this study focuses on hiding restrictive itemsets and lowering the side effects, effectively. This study proposes a novel sanitization method, which employs the hub and the authority weights of the web structure mining technique to hide the restrictive itemsets. Experimental results reveal that the proposed method has a low side effect.