Abstract:
Let an Abelian group (X,+) be the alphabet of R-round Markov block cipher with matrix P of transition probabilities of differentials; matrix size equals M=|X′|, X′=X∖{0}. Suppose spectrum of P satisfies the condition λ1=1>|λ2|>|λ3|⩾…⩾λM.
1. Extremal transition probabilities pab(R) and rows PR for a large number of rounds. Let P be diagonalizable: BPC=D=diag(1,λ2,…,λM), B=C−1, and there exist a,b∈X′ such that |Ca2|>|Ci2|, |B2b|>|Bjb| for all i≠a, j≠b. Then argmax(i,j)∈X′×X′|pij(R)−1M|=(a,b) and argmaxi∈X′|P(R)i−1M1|=a for all sufficiently large R, pab(R)−1M∼Ca2B2bλR2 and |Pa(R)−1M1|∼|Ca2||B2||λ2|R as R→∞.
2. Distinguishing attack by independent full codebooks. Let the cipher with alphabet X=Zn2 be Markovian (provided random uniformly distributed set of round keys k∼U(KR)) with matrix P, zi=zi(k) be the result of block i∈X transformation either by cipher (hypothesis H2) or random uniformly distributed substitution z(k) (hypothesis H1). Let (λ2,u) or (λ2,v) be left or right eigenpair of P, |u|=|v|=1, μ2(R)=uPRv↓≠0, S(k)=M∑{i,j}⊂Xuj−ivzj−zi. We prove that mean and variance of statistic S(k) equal 0 and M2M+12(M−2) respectively under hypothesis H1. If sets k(1),…,k(Nb)∼U(KR) are independent, Nb→∞, then for all 0<α<1 criterion d:S′(Nb)sign(μ2(R))>κ1−α√NMM−2⟹H2, where N=\dbinom{M+1}2 N_b, has error probability \alpha_1(d)\to\alpha. We show that \alpha_2(d)\approx \beta for large values of R and N_b\approx \frac{2(\kappa_{1-\alpha}+\kappa_{1-\beta})^2 }{(2^n \mu_2(R))^2}.
Keywords:
Markov block ciphers, distinguishing attack, matrix spectrum, transition probabilities of differentials, second dominant eigenvalue, independent full codebooks.
Bibliographic databases:
Document Type:
Article
UDC:519.23
Language: Russian
Citation:
O. V. Denisov, “Spectral probabilistic and statistical analysis of Markov ciphers”, Prikl. Diskr. Mat., 2021, no. 53, 12–31