可逆式隱私保護資料探勘之研究

A Study of Reversible Privacy-preserving Data Mining Technology

侯佩文(Pei-Wen Hou)

指導教授 ： 陳同孝

共同指導教授 ： 陳民枝

國立臺中科技大學/資訊學群/資訊工程系碩士班/碩士(2011年)

https://doi.org/10.6826/NUTC.2011.00017

摘要

隱私保護資料探勘(privacy-preserving data mining, PPDM)可有效避免資料庫遭有心人士以推理(inference)方式找出已保護的資料內容。但是，PPDM的不可還原特性，易造成關鍵決策欄位即為受保護欄位時，無法有效驗證資料庫之知識結果。故本研究保有PPDM保護隱私資料之特性並改善知識無法驗證之缺點，提出可逆式隱私保護資料探勘(reversible privacy-preserving data mining, RPPDM)，藉由可逆式資料隱藏技術擾動原始資料庫，並產生一把記錄RPPDM擾動規則之金鑰，交至合法使用者供還原隱私資料。同時，嵌入易碎型浮水印於擾動資料庫中，協助識別資料是否遭人竄改。
本文實驗中，以資料損失率、隱私外洩風險與分類器評估RPPDM擾動資料庫之成效。其實驗結果證明RPPDM演算可與原始知識結果相似，並有效控制隱私外洩風險於25%以下。所嵌入的易碎型浮水印也可加強擾動資料庫的防偽識別能力。

關鍵字

可逆式隱私保護資料探勘 ； 可逆式資料隱藏 ； 隱私保護資料探勘 ； 隱私資料安全

並列摘要

Privacy-preserving data mining (PPDM) may be used by perturbing or deleting information to protect the privacy data while preserving the analyzed knowledge to be similar to its original database. This method can effectively prevent interested parties from speculating the possibility of protected data by inference. However, the irreversible characteristic of PPDM resulted in ineffective verification of the knowledge in the database when critical decision fields were protected.
In this thesis, the PPDM characteristic of protecting privacy data is adopted with the proposed improved technique of reversible privacy-preserving data mining (RPPDM). RPPDM is applied to perturb the original database and a secret key was used to record the perturbation rule which may be used to retrieve the critical field. At same time, a fragile watermark was also embedded to help identify whether the information in the database had been tampered with.
Results were analyzed for information loss rate, discloser risk and classification of the perturbed database. Furthermore, the RPPDM algorithm can effectively control the risk of privacy leaks lower than 25% and the mined knowledge from the perturbed database resembles those of the original data. Also, the embedded fragile watermark increases the robustness of the perturbed database.

並列關鍵字

reversible privacy-preserving data mining ； reversible data hiding ； privacy-preserving data mining ； privacy data security

參考文獻

[1]R. Agrawal and R. Srikant, “Privacy-preserving Data Mining,” ACM SIGMOD Record, vol. 29, pp. 439-450, 2000.

[2]D. Agrawal and C. C. Aggarwal, “On the Design and Quantification of Privacy Preserving Data Mining Algorithms,” in Proceeding of 20th ACM SIGACT-SIGMOD-SIGART Symp. Principles of Database Systems, pp. 247-255, 2001

[3]A. M. Alattar, “Reversible Watermark Using Difference Expansion of Triplets,” in Proceedings of International Conference on Image Processing, pp. 501-504, 2003.

[4]C. Clifton, “Using Sample Size to Limit Exposure to Data Mining,” Journal of Computer Security, vol. 8, no. 8, pp. 281-307, 2000.

[5]D. Coltuc and J. M. Chassery, “Very Fast Watermarking by Reversible Contrast Mapping,” IEEE Signal Processing Letters, vol. 14, no. 4, pp. 255-258, 2007.