結合權限控制之高效能圖形結構隱私保護技術

Efficient and Effective Graph Structure Privacy Protection Technologies Combined with Access Control

曾詩翰(Shih-Han Tseng)

指導教授 ： 林真伊

國立臺中科技大學/資訊與流通學院/資訊管理系碩士班/碩士(2022年)

摘要

在現今的環境下，許多應用都會不斷的產生資料，比如：社交網路、物聯網、醫療健保等，為了創造更多的加值服務，通常會將這些資料進行公開，發布給研究學者或是專業人士進行分析，然而在資料中可能會存在著個人隱私資訊，倘若未將資料進行隱私保護處理，有心人士則可能會利用資料侵犯到用戶的隱私。
在過去資料隱私保護的應用中，多為處理表格結構資料，而近幾年，較複雜的圖形結構資料逐漸受到關注，當中，以k-anonymity為常見的隱私保護模式，許多學者發展出了不同的隱私保護演算法，使得至少有k筆資料在相同的等價類中，但這些演算法時常產生較高的資訊遺失情形與較差的執行時間效率，且匿名化後的圖形結構資料多為無法達成k-anonymity條件限制。而角色權限控管(RBAC)機制，可確保用戶僅能依照分配的角色來查詢授權的資料，有學者表示使用角色權限控管機制，在匿名化圖形結構資料進行查詢，可提升隱私保護的效用，然而過去相關的做法中，會導致角色查詢產生較高的不精確度。
「圖形結構泛化/聚類」是一種圖形結構資料進行匿名化隱私保護常見的方法，根據上述相關問題，本研究使用了這樣的方法，設計出兩種用於圖形結構資料的匿名化隱私保護演算法，並且結合了角色權限控管機制與k-anonymity隱私保護模式，進一步提升隱私保護的效用，而在匿名化過程中透過基於k-means技術來對圖形結構資料進行分群。根據實驗結果並與過去學者的作法相比，本研究所提出的演算法擁有較低資訊遺失程度，在執行時間效率上也有著顯著的改善，並且同時滿足了k-anonymity條件限制，也進一步的降低了整體角色查詢不精確度。

關鍵字

圖形結構 ； 隱私保護 ； 圖形結構泛化/聚類 ； 角色權限控管 ； k-anonymity ； k-means++ ； Constrained k-means

並列摘要

Nowadays, data are brought up constantly due to various applications, such as, social networks, Internet of Things, healthcare, and the like. To make more value-added services possible, those data would be publicly offered to researchers or professionals to do further analysis. However, personal information might be exposed at the same time, and lead to private data leaks. If the private data are not under proper protection, malicious actors can leverage them to violate the privacy of users.
Concerning the approaches of private data protection in the past, conducting table-structured data is commonly applied. Nevertheless, in recent years, dealing with more complicated graph-structured data has been attracting more attention, and k-anonymity is the most common privacy protection model applied. Many scholars have developed different algorithms of privacy protection, which are to make at least k-amount of data keep in the same equivalence class, but the risk of applying these algorithms is to undertake more frequent information loss and less efficiency. Graph-structured data also cannot mostly satisfy the condition of k-anonymity after they are anonymously applied. In contrast, Role-Based Access Control (RBAC) is to restrict the users only to inquire the authorized data in accordance with the assigned positions. Some scholars claim that by applying BBAC to do the inquires in the condition of anonymous graph-structured data, the privacy protection can be improved, but certain methods adopted in the past might lead to higher inaccuracy concerning role-based inquiries.
“Graph generalization/clustering” is the approach frequently adopted to apply to anonymous privacy protection upon graph-structured data. Regarding the problems mentioned above, in this paper, “graph generalization/clustering” is applied to figure out two ways of algorithms of anonymous privacy protection which is to apply to graph-structured data. These two algorithms are not only equipped with RBAC and k-anonymity modes of privacy protection, which is to upgrade the efficacy of privacy protection, but also clustering graph-structured data regarding k-means in the process of anonymity. According to the experiment results and comparing to the approaches scholars adopted before, the algorithms I proposed here are with less risk of information loss, significantly improving efficiency, and still satisfying the requirement of k-anonymity, which can lower the imprecision of role-based inquiries as well.

並列關鍵字

graph structure ； privacy protection ； graph generalization/clustering ； Role-Based Access Control ； k-anonymity ； k-means++ ； Constrained k-means

參考文獻

1. A. Labrinidis and H.V. Jagadish, Challenges and opportunities with big data, Proceedings of the VLDB Endowment, 2012.

2. S. Sagiroglu and D. Sinanc, Big data: A review, in 2013 International Conference on Collaboration Technologies and Systems (CTS), 2013.

3. J. Mondal and A. Deshpande, Managing large dynamic graphs efficiently, in Proceedings of the 2012 ACM SIGMOD International Conference on Management of Data, 2012.

4. S. Ji, P. Mittal, et al., Graph Data Anonymization, De-Anonymization Attacks, and De-Anonymizability Quantification: A Survey, IEEE Communications Surveys Tutorials, 2017.

5. A. Majeed and S. Lee, Anonymization Techniques for Privacy Preserving Data Publishing: A Comprehensive Survey, IEEE Access, 2021.