130 likes | 253 Vues
This presentation discusses a secure private matching scheme ensuring data privacy for both participants while achieving efficient computation of set intersections. It explores various adversary models, addressing semi-honest and malicious actors. Utilizing homomorphic encryption and balanced hashing, the proposed methods enable clients and servers to interact without revealing sensitive input data. The mechanics of oblivious transfer and polynomial interpolation are outlined, contributing to the overall robustness of the security guarantees. The findings are based on the original research presented at EUROCRYPT 2004 by Freedman, Nissim, and Pinkas.
E N D
Efficient Private Matching and Set Intersection (EUROCRYPT, 2004) Author:Michael J.Freedman Kobbi Nissim Benny Pinkas Presentered by Chia Jui Hsu Date:2009-02-10
Outline • Introduction • Private Matching Scheme • Adversary models • Security • Conclusion • References
Introduction (1/3) Intersection A B DataSets
Introduction (2/3) • Oblivious Transfer(忘卻式傳輸/模糊傳送) 模 糊 傳 送 OR Sender Receiver 1 out of 2 OT 1.傳送者不知道接收者是否得到密文 2.接收者只能得到他選擇的密文 M. Rabin, "How to Exchange Secrets by Oblivious Transfer", Technical Report TR-81,Aiken Computation Laboratory, Harvard Univ.,1981.
Introduction (3/3) • Homomorphic encryption system • E(m1)⊙E(m2)= E(m1 m2) • c=E(m), ck=E(km) Θ
Private Matching Scheme (1/4) • PM Scheme • client/chooser (C) and server/sender (S) • C inputs X = {x1,…,xkc} and S inputs Y = {y1,…,yks} • C learns X∩Y :PM(X,Y) • Polynomial input of size C 讓S算的變數
Private Matching Scheme (2/4) • Horner scheme • example • 若y=3,則P(y)=5
Private Matching Scheme (3/4) • 法二 • 法三 y=3,P(y)=5
Private Matching Scheme (4/4) Server Client X={x1,…xkc} Y={y1,…yks} 1.內插法算出多項式 2.對多項式的係數做同態加密 4.選擇一個亂數值γ 5. 3.上傳至Server 6.重新排列後回傳KS個 7.解密,若一樣,則解出y 不一樣,則解出亂數
Adversary models • Semi-honest • 1.pretecting the client • indistinguishability • 2.protecting the sender • comparison to the ideal model • Malicious • adversary may behave arbitrarily • 1.拒絕參與協定(PM) • 2.用任意值代替輸入 • 3.過早中止協定(PM)
Security • Correctness • C’s privacy is preserved • S’s privacy is preserved
Conclusion • use homomorphic encryption and balanced hashing for both semi-honest (standard model) and malicious (random oracle model) environments. • list length k, communication O(k), and computation is O(klnlnk).
References • Efficient Private Matching and Set Intersection, 2004 • http://en.wikipedia.org/wiki/Horner_scheme