差分隱私中的階層分群應用：減少雜訊添加與資訊損失;Applications of Hierarchical Grouping in Differential Privacy : Optimizing Noise Reduction and Information Loss.

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/98311

Please use this identifier to cite or link to this item: https://ir.lib.ncu.edu.tw/handle/987654321/98311

Title:	差分隱私中的階層分群應用：減少雜訊添加與資訊損失;Applications of Hierarchical Grouping in Differential Privacy : Optimizing Noise Reduction and Information Loss.
Authors:	林芊彤;Lin, Chien-Tung
Contributors:	資訊工程學系
Keywords:	差分隱私;階層分群;資訊損失;敏感度;資料隱私;Differential Privacy;Hierarchical Grouping;Information Loss;Sensitivity;Data Privacy
Date:	2025-07-22
Issue Date:	2025-10-17 12:37:17 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著資料應用的廣泛發展，如何在保護個人隱私的同時保留資料的分析價值，成為資料隱私保護領域的重要課題。差分隱私（Differential Privacy, DP）是目前主流的隱私保護框架，具備嚴謹的數學定義跟可量化的保障效果，但在實作上，為了達到預期的隱私強度，通常需要加入大量雜訊，導致資料效用明顯下降。為了降低雜訊造成的效用損失，許多研究聚焦於資料前處理的技術，如透過分群和聚合的方式降低查詢的敏感度，減少需要加入的雜訊。本研究提出一套結合階層分群和差分隱私的 HG-DP（Hierarchical Grouping for Differential Privacy）框架，允許使用者根據資料屬性所定義的階層結構進行分群，並在每一組內計算敏感數值的統計資訊。透過這樣的設計，不僅能提升分群策略的語意可解釋性與彈性，也能有效降低敏感度，減少需要加入的雜訊量。此外，我們也提供視覺化圖表，讓使用者能直觀比較不同階層組合在資訊損失與敏感度上的表現，作為選擇適合分組方式的參考。學生資料集的實驗結果顯示，HG-DP 框架具備兼顧隱私與資料效用的潛力。整體而言，本框架主要是協助使用者評估以及做出知情決策，分組方式仍由使用者自行判斷。;As data applications continue to grow, balancing privacy and utility has become a key challenge. Differential Privacy (DP) offers strong guarantees but often reduces utility due to added noise. To address this, we propose HG-DP (Hierarchical Grouping for Differential Privacy), a framework that groups data based on attribute hierarchies to lower sensitivity before applying DP. This improves both interpretability and utility. We also provide visualizations to help users compare grouping strategies by information loss and sensitivity. Experiments on a student dataset show that HG-DP supports privacy-preserving release while assisting user-driven decisions.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	7	View/Open

社群 sharing

Loading...