Cryptography: on the Hope for Privacy in a Digital World Omer Reingold VVeizmann and Harvard CRCS 1 So, is there Hope for Privacy? • No! Privacy is doomed! Enjoy your sandwiches … : Is this what we invited you for? • On second thought, the digital world gives new hope for privacy! – Selling digital goods (w/ Bill Aiello and Yuval Ishai) – Keyword database search (w/ Mike Freedman, Yuval Ishai, and Benny Pinkas) 2 Day to Day Breaches of Privacy • When/how can it be better? 3 Anonymity? Alice Not in this Talk! I can call you Betty, And Betty, when you call me, you can call me Al! Bob Call me Al ...... 4 Selling Digital Goods • How good are digital goods? – Entertainment: TV, music, video, books, software – Business: news, stock quotes, patents, layoff rumors – Research: papers, research databases, clip-art • What’s special about digital goods? – Typically of unlimited supply (easy to duplicate). – Easy to communicate and manipulate • Main goal: protect the privacy of clients – What – When – How much – (But not who) 5 Example Encrypted Individually ‘ ’, Vendor Buyer Key of 6 Oblivious Transfer (OT) [R], 1-out-of-N [EGL]: – X1 X2 X3 X4 Input: • Vendor: x1,x2,…,xn • Buyer: – … Xn j 1≤j≤n Output: • Vendor: nothing • Buyer: – xj Xj Privacy: • Vendor: learns nothing about j • Buyer: • learns nothing about xi for i ≠ j 4 – Not necessarily two messages – Related notions: Private Information Retrievable [CGKS] / Symmetrically- Private Information Retrievable [GIKM] 7 Priced OT [AIR] Vendor Buyer Initial payment $ b0 Set b=b0 Prices: p0=0, p1, p2 , … pn Buyer i ki Items: k0, k1, k2, … kn Vendor b← b - pi 8 Comparison with E-cash [Cha85,CFN88,...] E-cash Payment Goods Hides Access to goods Buyer Priced OT digital any who anonymous any digital what + any Vendor 9 General Perspective • Priced OT is an instance of secure two-party computation. • Theoretical plausibility result are known [Yao,GMW]. • However: General solutions are costly (computation, bandwidth, rounds). • A major endeavor in cryptography: Identifying interesting specific problems and suggesting more efficient solutions. 10 Tool: Homomorphic Encryption Plaintexts from (G,+) • E(a),E(b) E(a+b) E(a),c E(c·a) • |G| large prime • Can use either additive G=ZP or multiplicative GZ*P • In particular, can use El-Gamal. 11 Conditional Disclosure of Secrets [GIKM,AIR] E(q),pk Buyer (sk,pk) • • Vendor E(CDS( E(a) a ; V(q) )) a Honest Buyer: V(q) = True How to protect against a malicious Buyer? – Method 1: Buyer proves in ZK that V(q) = True; – Method 2: Vendor disclose a subject to the condition V(q) = True. • Notation: CDS( a ; V(q) ) 12 Conditional Disclosure of Secrets - Implementation E(q),pk Buyer (sk,pk) Vendor E(CDS( a ; V(q) )) a a,q,i G CDS(a ; q=i) : a+r(q-i) r R{1,…,|G|} E is homomorphic - E(CDS( a ; V(q) )) can be computed from E(q) • Information-theoretic security for Vendor (hides a). • Need to verify “validity” of pk; Easy for El-Gamal! 13 Application: 1-Round OT* [AIR,NP] Buyer Vendor E(q),pk q (sk,pk) x1 x2 xn E(CDS(x1 ; q =1)), … , E(CDS(xn ; q =n)) * Weakened / incomparable notion of security vs. simulation: • Vendor’s security: purely information-theoretic • Buyer’s security: privacy only. 14 Database Search • OT/PIR/SPIR allow to privately retrieve the ith entry of a database. Efficiency depends linearly (at least) on the size of the database. • Sometime this is not enough. For example, consider a list of fraudulent card numbers. A merchant wants to check if a particular number is in the least. • Use OT/PIR? – • Table of 1016 ≈ 253 entries, 1 if fraudulent, 0 otherwise? Works on supporting more general database search. 15 Keyword Search (KS): definition • Input: – – • Server: database X={ (xi,pi ) } , 1 ≤ i ≤ N • xi is a keyword (e.g. number of a corrupt card) • pi is the payload (e.g. why card is corrupt) Client: search word w (e.g. credit card number) Output: – Server: nothing – Client: • pi if i : xi = w Server: (x1,p1) (x2,p2) Client: w … (xn,pn ) Client output: (xj ,pj ) iff w=xj • otherwise nothing 16 Conclusions • Our expectation of privacy in the “digital world” should not be bounded to our “physical world” experiences. • The ability to duplicate, manipulate and communicate digital information is key. • Very powerful cryptographic tool in the form of secure function evaluation. • Research on efficient instantiations, possibly with some security relaxations. 17