Andrej Bogdanov Chinese University of Hong Kong ON
Andrej Bogdanov Chinese University of Hong Kong ON THE PROVABLE SECURITY OF HOMOMORPHIC ENCRYPTION based on joint work with Chin Ho Lee Northeastern Unversity Bertinoro Summer School | July 2014
Public-key bit encryption b Alice Enc. PK(b) PK Enc. PK(b) Bob b SK PK Dec. SK( message indistinguishability (PK, Enc. PK(0)) and (PK, Enc. PK(1)) are computationally indistinguishable )
El Gamal encryption g, h in some large cyclic group PK = ( g, h )such that g. SK = h r random Enc. PK(b) = ( gr, 2 bhrwhere ) Dec. SK(x, y) = b such that x. SK = 2 b y
Homomorphism of encryptions Enc. PK(b) = ( gr, 2 bhr ) strongly homomorphic Enc. PK(b) Enc. PK(b’) and Enc. PK(b + b’) are identically distributed weakly homomorphic Dec. SK(Enc. PK(b) Enc. PK(b’)) = b + b’
Does P ≠ NP imply cryptography? requires average-case hardness of distinguishing encryptions provided SAT is worst-case hard
Cryptography from lattices If short vectors in certain lattices are worst-case hard to find, then we have. . . one-way functions Ajtai but we can find them public-key encryption in NP ∩ co. NP Ajtai-Dwork “somewhat” homomorphic encryption Regev, Peikert, Gentry, Brakerski and Vaikutanathan, . . .
Reductions How to prove message indistinguishability? ? x ∈SAT q 1 a 1 q 2 a 2 YES/NO (PK, Enc. PK(b)) distinguisher biased towards b
From reductions to proof systems L R distinguisher verifier prover randomness for R transcript is it correct? for every query (PK, C) answer b are they correct? randomness r s. t. Enc. PK(b, r) = C OK Brassard
From reductions to proof systems Conclusion A reduction from L to distinguishing Enc implies that L is in NP ∩ co. NP Yes, but under implicit assumption that queries always have a unique answer Goldreich and Goldwasser
Brassard’s assumption for every PK what if Enc. PK(1) Enc. PK(0) query Enc. PK(1) Enc. PK(0)
Restricting the reduction For general encryptions, best we can say If reduction is nonadaptive then L is in AM ∩ co. AM Feigenbaum and Fortnow, B. and Trevisan, Akavia Goldreich Goldwasser and Moshkovitz
Our result Let f be a “polynomially sensitive” function If Enc has weak homomorphic evaluator f, then L is in AM ∩ co. AM Reduction can be adaptive, queries arbitrary If reduction has constant query complexity, then L is in statistical zero-knowledge
Sensitivity of functions f: 0 1100 1 0110 0 1 0101 sens 0 f(0100) = 2 sens 0 f = maxx sens 0 f(x) 0100 f: {0, 1}n → {0, 1} is polynomially sensitive if sens 0 f, sens 1 f are at least n. W(1)
previous works AM Arbitrary encryptions, nonadaptive reductions M A co Homomorphic encryptions, arbitrary reductions SAT SZK P Homomorphic encryptions, reductions of constant query complexity
Rerandomization The ability to map a ciphertext into an i. i. d ciphertext without knowing the secret key El Gamal example PK = ( g, h ) such that g. SK = h C = ( g r, 2 b h r ) Rer. PK(C) = C ∙ ( gr’, hr’ ) is i. i. d with C
Enc(1) Enc(b) Enc(0) Rerandomization from evaluation H Enc(b) strong homomorphic evaluator for majority
Enc(0) Rerandomization from evaluation H To H, Enc(0) indistinguishable from Enc(0) so output of H must forget most of Enc(0)
Rerandomization from evaluation Lemma If H is a strong homomorphic evaluator for majority on k bits, then (Enc(b), Rer(Enc(b)) is √c/k-close to a pair of independent encryptions of b. We prove a weaker version for weak homomorphic evaluators and any sensitive f.
Distinguishing rerandomizations Encryption can be broken using rerandomization and an SZK oracle Rer(Enc(b) ) vs. Enc(0) If b = 0, they are statistically close If b = 1, they must be statistically far so they can be distinguished in SZK
The rest of the proof Since we can decrypt in SZK, L can be solved with reduction + SZK oracle So L is in BPPSZK ⊆ AM ⋂ co. AM Mahmoody and Xiao For weak homomorphism and general f, not sure if true; we give new proof system
Quality of rerandomization Lemma If H is a homomorphic evaluator for majority on k bits, then (Enc(b), Rer(Enc(b)) is √c/k-close to a pair of independent encryptions of b. For strong homomorphic evaluation, we can make this exponentially small.
Enc(1) Enc(b) Enc(0) Improving the rerandomization H Enc(b) Algorithm: Apply H iteratively t times. Enc(1) H Enc(b) Enc(0)
H Enc(b) H Enc(1) H Enc(b) Enc(0) Enc(1) Enc(1) Enc(b) Enc(0) Analysis H Enc(0)
H Enc(1) Enc(0) Enc(1) Enc(0) Analysis H Enc(0)
Analysis If we recurse t times, original Enc(b) could be any one of 2 t inputs Applying lemma, distinguishing advantage drops to O(√c/2 t) Value of t is determined by quality of H Statistical distance between output of H and actual encryption
Rerandomization theorem f : any function except for AND, OR, NOT Assume f has strong homomorphic evaluator with quality 2 -h then there is a rerandomization with statistical error 2 -W(h).
- Slides: 26