on the circular security of bit encryption
play

On the Circular Security of Bit Encryption Ron Rothblum Weizmann - PowerPoint PPT Presentation

On the Circular Security of Bit Encryption Ron Rothblum Weizmann Institute Circular Security Circular Security Q: Is it in general safe to encrypt your own key? A: For some schemes (e.g. [BHHO08,ACPS09] ) yes but in general No! Circular


  1. On the Circular Security of Bit Encryption Ron Rothblum Weizmann Institute

  2. Circular Security

  3. Circular Security Q: Is it in general safe to encrypt your own key? A: For some schemes (e.g. [BHHO08,ACPS09] ) yes but in general No!

  4. Circular Security

  5. Public Key Example Public Key Example

  6. Circular Security of Bit Encryption Since general case is false, focus on interesting special case of bit-encryption . Why bit-encryption? 1. Most candidate FHE are bit-encryption whose semantic-security relies on their circular security (which is not understood). 2. Seems most natural way to foil the previous counterexample and get circular security for “free”.

  7. Bit-Encryption Conjecture Conjecture: [Folklore] Every semantically-secure bit-encryption scheme is circular secure. Focus of this work is showing obstacles to proving the conjecture.

  8. Our Results 1. A scheme that is circular insecure but is semantically secure based on multilinear maps. 2. Cannot prove the conjecture via a blackbox reduction. 3. Equivalence of different security notions for circular security of bit- encryption.

  9. Our Results 1. A scheme that is circular insecure but is semantically secure based on multilinear maps. 2. Cannot prove the conjecture via a blackbox reduction. 3. Equivalence of different security notions for circular security of bit- encryption.

  10. Our Assumption An extension of an assumption made on groups with bilinear maps to groups with multilinear maps.

  11. Multilinear Maps

  12. Multilinear Maps There exist trivial multilinear maps unconditionally but for crypto, need computational problems such as discrete-log to be hard. Do there exist multilinear groups on which discrete-log (and friends) are hard? [BS03]

  13. (Silly) Example (Silly) Example

  14. SXDH Assumption [BGMM05, ACHM05] c

  15. Theorem

  16. El-Gamal Variant Key Generation: Decrypt(c,d):

  17. Our Scheme Key Generation:

  18. Our Scheme Key Generation:

  19. Circular Security Attack … … … … … … …

  20. Circular Security Attack …. …

  21. Circular Security Attack …. …

  22. Circular Security Attack …. …

  23. Circular Security Attack With overwhelming probability

  24. Our Results 1. A scheme that is circular insecure but is semantically secure based on multilinear maps. 2. Cannot prove the conjecture via a blackbox reduction. 3. Equivalence of different security notions for circular security of bit- encryption.

  25. Blackbox Impossibility Result No blackbox reduction from circular-security of bit-encryption scheme to semantic-security (or even CCA security) of the same scheme. Blackbox access to encryption-scheme and adversary. Incomparable to [HH09] KDM blackbox separation.

  26. [HH09] KDM Blackbox Impossibility [HH09] KDM Blackbox Impossibility

  27. A Blackbox Reduction A Blackbox Reduction Encryption Circular Security Circular Security Scheme Scheme Adversary Challenger Challenger Reduction (Semantic Security (Semantic Security Adversary)

  28. Our Results 1. A scheme that is circular insecure but is semantically secure based on multilinear maps. 2. Cannot prove the conjecture via a blackbox reduction. 3. Equivalence of different security notions for circular security of bit- encryption.

  29. Circular Security Definitions

  30. Equivalence Result

  31. Open Problems

  32. Thank you!

Recommend


More recommend