The Cut-and-Choose Game and its Application to Cryptographic Protocols
Ruiyu Zhu, Yan Huang, Jonathan Katz, abhi shelat
Indiana University Northeastern University
- U. Maryland
The Cut-and-Choose Game and its Application to Cryptographic - - PowerPoint PPT Presentation
The Cut-and-Choose Game and its Application to Cryptographic Protocols Ruiyu Zhu, Yan Huang, Jonathan Katz, abhi shelat Northeastern Indiana University U. Maryland University What is Cut-and-Choose What is Cut-and-Choose Applications of
The Cut-and-Choose Game and its Application to Cryptographic Protocols
Ruiyu Zhu, Yan Huang, Jonathan Katz, abhi shelat
Indiana University Northeastern University
What is Cut-and-Choose
What is Cut-and-Choose
Applications of Cut-and-Choose
–
–
–
–
BBSU, FC 12 Blum, ICM 86 CKV, Crypto 10 Lindell, Crypto 13 HKE, Crypto13 AMPR, Crypto14 Brandão, AsiaCrypt 13 SS, EuroCrypt 11 LP, Eurocrypt 07
Applications of Cut-and-Choose
–
–
–
–
BBSU, FC 12 Blum, ICM 86 CKV, Crypto 10 Lindell, Crypto 13 HKE, Crypto13 AMPR, Crypto14 Brandão, AsiaCrypt 13 SS, EuroCrypt 11 LP, Eurocrypt 07
Cut-and-Choose in Secure Computation
Eval Chk Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits
Three Flavors of Cut-and-choose
– Secure if at least one evaluation-circuit is correct.
– Secure if the majority of evaluation-circuits are correct.
– Amortizing cost over multiple executions.
Lindell, Crypto 13 HKE, Crypto 13 AMPR, Crypto 14 Brandão, AsiaCrypt 13 LR, Crypto 14 NO, TCC09 FJN+, EuroCrypt13 LP, EuroCrypt07 LP, JoP12 LP, SCN 08 Woodruff, EuroCrypt 07 SS’ EuriCrypto 11
Three Flavors of Cut-and-choose
– Secure if at least one evaluation-circuit is correct.
– Secure if the majority of evaluation-circuits are correct.
– Amortizing cost over multiple executions.
Lindell, Crypto 13 HKE, Crypto 13 AMPR, Crypto 14 Brandão, AsiaCrypt 13 LR, Crypto 14 NO, TCC09 FJN+, EuroCrypt13 LP, EuroCrypt07 LP, JoP12 LP, SCN 08 Woodruff, EuroCrypt 07 SS’ EuriCrypto 11
Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits
Existing SingleCut Strategy
Expected cost: checking cost × 𝑡 2 + evaluation cost× 𝑡 2
Lindell, Crypto 13
Eval Chk Eval Chk Eval 𝑡:the security parameter
Garbled Circuit Garbled Circuit
The Cost Gap
Seed Hash
Checking Evaluation Time Cost Ratio
2 ~ 30
Bandwidth Cost Ratio 107~108
16 bytes 32 bytes
Our Key Intuition
Evaluate less and check more. Use mixed-strategies: determine the number
custom distribution. Use linear programming to find optimal parameters.
Problem Formulation
Want to minimize
𝜁 Upper-bound on the security failure rate 𝑠 Cost ratio 𝑇<=> Generator’s strategy 𝑇=?@A Evaluator’s strategy
𝔽[cost(𝑠, 𝑇=?@A)] Subject to: Prfailure 𝑇=?@A, 𝑇<=> ≤ 𝜁, ∀𝑇<=>
“For all cheating strategies”
Prfailure 𝑇=?@A, 𝑇<=> ≤ 𝜁, ∀𝑇<=>
Problem Formulation
Subject to: Want to minimize
𝜁 Upper-bound on the security failure rate 𝑠 Cost ratio 𝑻𝒉𝒇𝒐 Generator’s strategy 𝑻𝒇𝒘𝒃𝒎 Evaluator’s strategy
𝔽[cost(𝑠, 𝑇=?@A)]
“For all cheating strategies”
𝑇<=> and 𝑇=?@A in SingleCut
𝑜 The total number of circuits 𝑇<=> A random variable over {0,1}> 𝑇=?@A A random variable over {0,1}>
𝑇<=> and 𝑇=?@A in SingleCut
Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits Garbled Circuits 1 1 1 1 1 1 Eval Chk Eval Chk Eval My only choices are which circuits to form improperly. I could map between binary string and strategy So could I
Failure: 𝑇<=> = 𝑇=?@A
𝔽[cost 𝑠, 𝑇=?@A ] = U(𝑗𝑠 + 𝑜 − 𝑗 X 1)𝑦Z
> Z[\
Expected Cost of SingleCut
𝑜 Total number of circuits 𝑦Z Probability of evaluating 𝑗 circuits
𝔽[cost 𝑠, 𝑇=?@A ] = U(𝑗𝑠 + 𝑜 − 𝑗 X 1)𝑦Z
> Z[\
𝔽[cost 𝑠, 𝑇=?@A ] = U(𝑗𝑠 + 𝑜 − 𝑗 X 1)𝑦Z
> Z[\
𝔽[cost 𝑠, 𝑇=?@A ] = U(𝑗𝑠 + 𝑜 − 𝑗 X 1)𝑦Z
> Z[\
𝔽[cost 𝑠, 𝑇=?@A ] = U(𝑗𝑠 + 𝑜 − 𝑗 X 1)𝑦Z
> Z[\
# of circuits to evaluate Total # of circuits
Constraints on 𝑦Z (because it’s a probability distribution)
> Z[\
𝑜 Total number of circuits 𝑦Z Probability of evaluating 𝑗 circuits
Prfailure 𝑇=?@A, 𝑇<=> ≤ 𝜁
Security Holds ∀ 𝑇<=>, Pr 𝑇=?@A = 𝑇<=> ≤ 𝜁 ∀𝑏 ∈ 0,1 >, Pr(𝑇=?@A = 𝑏) ≤ 𝜁
Probability that evaluator picks any SPECIFIC strategy a is bounded by 𝜁.
There are >
Z pure strategies
that evaluate 𝑗 circuits.
∀𝑏 ∈ 0,1 >, Pr (𝑇=?@A = 𝑏) ≤ 𝜁
Each pure strategy can be picked with probability at most 𝜁.
Recap
Subject to:
U(𝑗𝑠 + 𝑜 − 𝑗)𝑦Z
> Z[\
𝑦Z ≤ 𝜁 𝑜 𝑗 𝑦Z ≥ 0 U 𝑦Z
> Z[\
= 1
Minimize:
Fractional Knapsack Problem
Capacity: 1/𝜁 units
Unit Cost: 𝑜 𝑠 + 𝑜 − 1 2𝑠 + 𝑜 − 2 𝑜𝑠 Units 𝑜 𝑜 1 𝑜 2 𝑜 𝑜
A greedy algorithm solves it in linear time.
Find the Best 𝑜
to find the one with minimal cost.
https://github.com/Opt-Cut-N-Choose
Required by the security parameter 𝜁 Achievable with the SingleCut strategy of [Lindell, Crypto13].
logd 1 𝜁 𝑠 + 1 2 logd 1 𝜁
Range of 𝑜
Sample SingleCut Strategy for AES
Our technique 𝒐 = 𝟑𝟑𝟕𝟖 𝒋 𝒚𝒋 as % 9.09 X 10lmm 1 2.06 X 10lo 2 2.34 X 10l7 3 1.77 X 10lm 4 99.8 X 10\ Save 77.5% b/w Classical Strategy 𝒐 = 𝟓𝟏 𝒋 𝒚𝒋 as % 9.09 X 10lmm ⋯ ⋯ 19 11.9 X 10\ 20 12.5 X 10\ ⋯ ⋯ 40 9.09 X 10lmm
Bandwidth cost ratio: 𝑠 = 4533 For AES
Improvements on SingleCut
0% 20% 40% 60% 80%
Savings Cost Ratio r
Savings=1− costthis work costbest prior work
100 101 102 103 104
Improvements on SingleCut
0% 20% 40% 60% 80%
Savings Cost Ratio r
100 101 102 103 104
AES fp-multiply
Formulation for MajorityCut
See the paper for details.
Subject to:
U(𝑗𝑠 + 𝑜 − 𝑗)𝑦Z
> Z[\
U 𝑦Z X 𝑜 − 𝑐 𝑗 − 𝑐 / 𝑜 𝑗
}~• (>,d€) Z[€
≤ 𝜁 𝑦Z ≥ 0 U 𝑦Z
> Z[\
= 1
Minimize:
Sample MajorityCut Strategy
Our technique 𝒐 = 𝟐𝟖𝟔 𝒋 𝒚𝒋 as % 𝒋 𝒚𝒋 as % 7 1 X 10l7 17 1.23 9 9 X 10l7 19 5.36 11 7 X 10lƒ 21 20.9 13 4.54 X 10ld 23 72.2 15 0.25 Save 26.6% time
Time cost ratio: 𝑠 = 10
Classical Strategy 𝒐 = 𝟐𝟑𝟓 𝒋 𝒚𝒋 as % 43 100
Improvements on MajorityCut
Savings=1− costthis work costbest prior work
Savings Cost ratio r
100 102 104 106 108 0% 20% 40% 60% 80% 100%
Improvements on MajorityCut
Savings Cost ratio r
100 102 104 106 108 0% 20% 40% 60% 80% 100%
fp-multiply AES
Improvements on BatchedCut
Savings=1− costthis work costbest prior work
Savings Cost ratio r
N=100 N=200 N=10000
N is the size
100 101 102 103 104 105 0% 10% 20% 30% 40% 50%
Conclusion
The game solvers are available at https://github.com/cut-n-choose.
Cut-and-choose protocols should be appropriately configured based on the security requirement and the cost ratio benchmarked at run-time.
Ruiyu Zhu: zhu52@indiana.edu