1. On the Practical Feasibility of Secure Distributed Computing A Case Study Gregory Neven, Frank Piessens, Bart De Decker Dept. of Computer Science, K.U.Leuven Celestijnenlaan 200A, B-3001 Heverlee, Belgium

2. 2 Secure Distributed Computing Given –n different participants P 1 …P n –each participant P i has a secret input x i –some function f How to –compute y = f(x 1,…, x n ) –without P i being able to learn anything more about x j (i  j) than what is implied by the function result itself Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

3. 3 Secure Distributed Computing (cont) Practical applications –Second price auctions –Voting –Privacy for mobile code –Secret Query Database Query a database while preserving privacy of query Example Alice sells records from database of CV’s Bob doesn’t want just any CV doesn’t want to reveal his selection criteria Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

4. 4 Overview Secure Distributed Computing –Trivial solution –Protocol (by Abadi & Feigenbaum) –Other protocols Case study: Secret Query Database –Implementation –Assessment Conclusion

5. 5 Trivial Solution Using a Trusted Third Party (TTP) x1x1 x2x2 xnxn y y y y = f(x 1,…,x n ) Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

6. 6 Outline of the Protocol Two participants –Alice knows secret data x = x 1 x 2 … –Bob knows secret function f (as a boolean circuit) –Compute y = f(x) without compromising their secrets Outline Alice Bob x = x 1 x 2 … E n (y 1 ), E n (y 2 ), … f y y Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

7. 7 Encryption Scheme Probabilistic encryption –one plaintext  many possible ciphertexts –secure for small message spaces –disadvantage: huge data blowup – Homomorphic encryption scheme –E(x) op E(y) = E(x op’ y) –Properties: Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

8. 8 Evaluation of the Circuit NOT-gate XOR-gate AND-gate ? No interaction with Alice! Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

9. 9 Evaluation of the Circuit (cont) AliceBob ? Choose random c 1, c 2 Decryption  d 1, d 2 !  communication overhead  Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

10. 10 Other SDC Protocols Goldreich, Micali and Wigderson (1987) –Two-party –Based on symmetric encryption and oblivious transfer Sander and Tschudin (1998) –Two-party –Autonomous protocol –Based on dual-homomorphic encryption schemes –Polynomial evaluation only Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

11. 11 Chaum, Damgard and van de Graaf (1988) –Multi-party –Based on blindable bit commitments Franklin and Haber (1996) –Multi-party –Based on group-oriented cryptography Other SDC Protocols (cont) Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

12. 12 Secret Query Database Problem statement –Query DB while preserving privacy of query Example –Alice sells records from database of CV’s = secret data x –Bob doesn’t want just any CV doesn’t want to reveal his selection criteria = secret function Q( ) Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

13. 13 Implementation AliceBob n = pq record x n, E n (x 1 ), E n (x 2 ),… query Q evaluation E n (Q(x)) Decrypt  Q(x) Q(x) = 1 ? x Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

14. 14 Security Trade-Off Security parameter: |n| –Each record different n  512 bits Huge data blowup! –1 plaintext bit  512 encrypted bits Encrypted records reusable –p and q are never revealed –Same encryption used for multiple session  Edited on CD-ROM Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

15. 15 Assessment Typical values –|record x| = 500 bytes –|database| = 1000 records –|query Q| = 1000 gates –|n| = 512 bits (security parameter) Communication complexity Per recordTotal On CD-ROM250 KB244 MB Over network94 KB92 MB Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

16. 16 Conclusion High overhead, but –increasing bandwidth of the Internet –trade-off communication  security –trade-off communication  query complexity –mobile agent technology SDC is ready for practical applications Introduction Secure Distributed Computing Case Study: Secret Query Database Conclusion

17. 17 Quadratic Residuosity Suppose –p and q primes congruent to 3 mod 4 –n = pq – a is a quadratic residue (QR) mod n iff Quadratic Residuosity Assumption (QRA): Is a a QR mod n or not? –easy if p and q are known –hard if p and q are unknown

18. 18 Some Properties Inversion If a is a QR mod n, then is a QNR mod n (and vice versa) Multiplication mod n ab a  b mod n QR QNR QRQNR QR ab 000 011 101 110 a  b

19. 19 Efficiency Improvement AliceBob ? Choose random c 1, c 2 Decryption  d 1, d 2 !  If c 2 = 0  b 1  c 2 = 0  E n (b 1  c 2 ) = E n (0) If c 2 = 1  b 1  c 2 = b 1  E n (b 1  c 2 ) = E n (b 1 )