1 In what follows, we consider the Boolean algebra on the set
2 $\Bool=\{0,1\}$ with the classical operators of conjunction '.',
3 of disjunction '+', of negation '$\overline{~}$', and of
4 disjunctive union $\oplus$.
6 Let $n$ be a positive integer. A {\emph{Boolean map} $f$ is
7 a function from $\Bool^n$
10 $x=(x_1,\dots,x_n)$ maps to $f(x)=(f_1(x),\dots,f_n(x))$.
11 Functions are iterated as follows.
12 At the $t^{th}$ iteration, only the $s_{t}-$th component is
13 ``iterated'', where $s = \left(s_t\right)_{t \in \mathds{N}}$ is a sequence of indices taken in $\llbracket 1;n \rrbracket$ called ``strategy''. Formally,
14 let $F_f: \llbracket1;n\rrbracket\times \Bool^{n}$ to $\Bool^n$ be defined by
16 F_f(i,x)=(x_1,\dots,x_{i-1},f_i(x),x_{i+1},\dots,x_n).
18 Then, let $x^0\in\Bool^n$ be an initial configuration
20 \llbracket1;n\rrbracket^\Nats$ be a strategy,
21 the dynamics are described by the recurrence
22 \begin{equation}\label{eq:asyn}
27 Let be given a Boolean map $f$. Its associated
28 {\emph{iteration graph}} $\Gamma(f)$ is the
29 directed graph such that the set of vertices is
30 $\Bool^n$, and for all $x\in\Bool^n$ and $i\in \llbracket1;n\rrbracket$,
31 the graph $\Gamma(f)$ contains an arc from $x$ to $F_f(i,x)$.
34 Let us consider for instance $n=3$.
36 $f^*: \Bool^3 \rightarrow \Bool^3$ be defined by
38 (x_2 \oplus x_3, \overline{x_1}\overline{x_3} + x_1\overline{x_2},
39 \overline{x_1}\overline{x_3} + x_1x_2)$.
40 The iteration graph $\Gamma(f^*)$ of this function is given in
41 Figure~\ref{fig:iteration:f*}.
46 \includegraphics[scale=0.5]{images/iter_f0c}
49 \caption{Iteration Graph $\Gamma(f^*)$ of the function $f^*$}\label{fig:iteration:f*}
54 % It is easy to associate a Markov Matrix $M$ to such a graph $G(f)$
57 % $M_{ij} = \frac{1}{n}$ if there is an edge from $i$ to $j$ in $\Gamma(f)$ and $i \neq j$; $M_{ii} = 1 - \sum\limits_{j=1, j\neq i}^n M_{ij}$; and $M_{ij} = 0$ otherwise.
60 % The Markov matrix associated to the function $f^*$ is
63 % M=\dfrac{1}{3} \left(
64 % \begin{array}{llllllll}
79 It is usual to check whether rows of such kind of matrices
80 converge to a specific
82 Let us first recall the \emph{Total Variation} distance $\tv{\pi-\mu}$,
83 which is defined for two distributions $\pi$ and $\mu$ on the same set
85 $$\tv{\pi-\mu}=\max_{A\subset \Omega} |\pi(A)-\mu(A)|.$$
87 % $$\tv{\pi-\mu}=\frac{1}{2}\sum_{x\in\Omega}|\pi(x)-\mu(x)|.$$
89 Let then $M(x,\cdot)$ be the
90 distribution induced by the $x$-th row of $M$. If the Markov chain
92 $M$ has a stationary distribution $\pi$, then we define
93 $$d(t)=\max_{x\in\Omega}\tv{M^t(x,\cdot)-\pi}.$$
94 Intuitively $d(t)$ is the largest deviation between
95 the distribution $\pi$ and $M^t(x,\cdot)$, which
96 is the result of iterating $t$ times the function.
97 Finally, let $\varepsilon$ be a positive number, the \emph{mixing time}
98 with respect to $\varepsilon$ is given by
99 $$t_{\rm mix}(\varepsilon)=\min\{t \mid d(t)\leq \varepsilon\}.$$
100 It defines the smallest iteration number
101 that is sufficient to obtain a deviation lesser than $\varepsilon$.
102 % Notice that the upper and lower bounds of mixing times cannot
103 % directly be computed with eigenvalues formulae as expressed
104 % in~\cite[Chap. 12]{LevinPeresWilmer2006}. The authors of this latter work
105 % only consider reversible Markov matrices whereas we do no restrict our
106 % matrices to such a form.
110 Let us finally present the pseudorandom number generator $\chi_{\textit{14Secrypt}}$
111 which is based on random walks in $\Gamma(f)$.
112 More precisely, let be given a Boolean map $f:\Bool^n \rightarrow \Bool^n$,
113 a PRNG \textit{Random},
114 an integer $b$ that corresponds to an awaited mixing time, and
115 an initial configuration $x^0$.
116 Starting from $x^0$, the algorithm repeats $b$ times
117 a random choice of which edge to follow and traverses this edge.
118 The final configuration is thus outputted.
119 This PRNG is formalized in Algorithm~\ref{CI Algorithm}.
124 \begin{algorithm}[ht]
126 \KwIn{a function $f$, an iteration number $b$, an initial configuration $x^0$ ($n$ bits)}
127 \KwOut{a configuration $x$ ($n$ bits)}
129 \For{$i=0,\dots,b-1$}
131 $s\leftarrow{\textit{Random}(n)}$\;
132 $x\leftarrow{F_f(s,x)}$\;
136 \caption{Pseudo Code of the $\chi_{\textit{14Secrypt}}$ PRNG}
140 This PRNG is a particularized version of Algorithm given in~\cite{BCGR11}.
141 Compared to this latter, the length of the random
142 walk of our algorithm is always constant (and is equal to $b$) whereas it
143 was given by a second PRNG in this latter.
144 However, all the theoretical results that are given in~\cite{BCGR11} remain
145 true since the proofs do not rely on this fact.
147 Let $f: \Bool^{n} \rightarrow \Bool^{n}$.
148 It has been shown~\cite[Th. 4, p. 135]{BCGR11}} that
149 if its iteration graph is strongly connected, then
150 the output of $\chi_{\textit{14Secrypt}}$ follows
151 a law that tends to the uniform distribution
152 if and only if its Markov matrix is a doubly stochastic matrix.
154 Let us now present a method to
156 with Doubly Stochastic matrix and Strongly Connected iteration graph,
157 denoted as DSSC matrix.