對後端資訊系統而言，目前對個人資料保護的要求與手段，是將機敏性資料隔絕使
一般大眾不能直接取得。至少就我們近來接觸到的高教校務研究，乃至一般公家機關的
公開資料，對於間接由一些非敏感性公開資料，拼湊推論而推論出機敏性資料的可能，
一直並沒有討論與面對。
在美國雖有「馬賽克理論」與相關案例，對間接推論行為有訂定資料保護權責，但
保護手段並沒有積極研究。若長此以往不面對處理，可能抑制資訊保有者彼此整合交流
與公開，使資訊片面破碎，而社會空有豐富資訊卻不能有意義的全盤了解與分析。
本研究探索功能相依性，據此得出能推論機敏資料的高風險屬性集合，並藉由這些
高風險屬性集合比對使用者在查詢時是否存在機敏資料推論行為，從而加以防範保護。

From the back-end data system point of view, the primary personal information protection
mechanism is to block the direct accessing of sensitive data. We have observed the related issues
in fields of Institutional Research, as well as governments’ information publication. And the
possibility that sensitive data may be indirectly inferenced by public information, have not been
addressed.
In United States, there are cases and discussions about “Mosaic theory”. And
responsibilities of data holders were legally stated. But no known researches were invested to
create a responsible mechanism. This may lead to a situation where data holders will not
willingly integrate, exchange, and publish their data. Our society may not be able to
comprehensively understand ourselves and conduct effective analysis, even though we do have
huge volume oh data.
This research explores the functional dependencies, and compute risky column sets based
on them. We can then process users’ queries and initiate protection operation if risky data are
involved.

〔1〕Kerr, O. S. (2012, 4). The Mosaic Theory of the Fourth Amendment. 111 Michigan Law Review 311 (2012)., p. 44.
〔2〕United States v. Jones。取自網路http://volokh.com/2012/03/18/more-cases-on-the-mosaic-theory-and-the-implications-of-jones/
〔3〕Mannila, H., & Räihä, K.-J. (1994, 2). Algorithms for inferring functional dependencies from relations. Data & Knowledge Engineering, pp. 83-99.
〔4〕去識別化。取自網路https://nvlpubs.nist.gov/nistpubs/ir/2015/NIST.IR.8053.pdf
〔5〕Lattice取自網路。https://en.wikipedia.org/wiki/Lattice_(order)
〔6〕Papenbrock, T., Ehrlich, J., & Marten, J. (2015, June). Functional dependency discovery: an experimental evaluation of seven algorithms. The Proceedings of the VLDB Endowment (PVLDB), pp. 1082-1093.
〔7〕Huhtala, Y., Kärkkäinen, J., Porkka, P., & Toivonen, H. (1999, Jan). Tane: An Efficient Algorithm for Discovering Functional and Approximate Dependencies. The Computer Journal, 42(2), pp. 100 - 111.
 
〔8〕Kivinen, J., & Mannila, H. (2005). Approximate dependency inference from relations. International Conference on Database Theory (pp. 86-98). Berlin: Heidelberg.
〔9〕Kivinen, J., & Mannila, H. (1995). Approximate inference of functional dependencies from relations. Theoretical Computer Science (pp. 129-149). Elsevier B.V.
〔10〕Abedjan, Z., Schulze, P., & Naumann, F. (2014). DFD: Efficient Functional Dependency Discovery. Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management (pp. 949-958). Shanghai, China: ACM.
〔11〕cardinality。取自網路https://en.wikipedia.org/wiki/Cardinality https://en.wikipedia.org/wiki/Cardinality_(SQL_statements)
〔12〕cardinal numbers。取自網路https://en.wikipedia.org/wiki/Cardinal_number
〔13〕entropy。取自網路https://zh.wikipedia.org/wiki/%E7%86%B5_(%E4%BF%A1%E6%81%AF%E8%AE%BA)
〔14〕Quinlan, J. R. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993.
〔15〕個人資料保護法。取自網路http://www.kmh.moj.gov.tw/ct.asp?xItem=195519&ctNode=5277&mp=008
〔16〕校務研究規範。取自網路http://ir.ncu.edu.tw/images/%E6%A0%A1%E5%8B%99%E7%A0%94%E7%A9%B6%E8%B3%87%E6%96%99%E7%B3%BB%E7%B5%B1%E5%BB%BA%E7%BD%AE%E5%8F%8A%E4%BD%BF%E7%94%A8%E4%BD%9C%E6%A5%AD%E8%A6%81%E9%BB%9E%E8%88%87%E6%B5%81%E7%A8%8B%E5%9C%961071001.pdf?fbclid=IwAR23QFcA9hbP7c3z5bkNbBcawiNLlts3wlVAO0KregwdpbhiBHwG49L_E8I
〔17〕Bertino, E., Byun, J.-W., & Li, N. (2004). Privacy-Preserving Database Systems. International School on Foundations of Security Analysis and Design (pp. 178-206). Berlin: Heidelberg.
〔18〕Olumofin, F., & Goldberg, I. (2010). Privacy-Preserving Queries over Relational Databases. International Symposium on Privacy Enhancing Technologies Symposium (pp. 75-92). Berlin: Heidelberg.
〔19〕Apriori演算法。取自網路https://en.wikipedia.org/wiki/Apriori_algorithm