On the nonlinearity of linear recurrence sequences (Q2488716)

From MaRDI portal
scientific article
Language Label Description Also known as
English
On the nonlinearity of linear recurrence sequences
scientific article

    Statements

    On the nonlinearity of linear recurrence sequences (English)
    0 references
    0 references
    0 references
    11 May 2006
    0 references
    Let \(p\) be a prime number and \({\mathbb F}_p\) the finite field of \(p\) elements identified with the set \(\{0,1,\ldots, p-1\}\). Let \((s_k)_{k=0}^\infty \) be a linear recurrence of order \(n\) over \({\mathbb F}_p\) and suppose its characteristic polynomial is irreducible. Let \({\mathcal B}\) denote the \(n\)-dimensonal cube \({\mathbb F}_p^n\) and introduce the inner product \(\langle h,r\rangle=h_1r_1+\cdots+h_nr_n\) for \(h=(h_1,\ldots,h_n)\) and \(r=(r_1,\ldots,r_n)\) in \(\mathcal B\). The main result of the paper is an estimate for the exponential sum \[ T(b)=\sum_{k=0}^{t-1} e_p(s_k+\langle b,k\rangle) \] related to the discrete Fourier transform, where \(e_p(z)=e^{2\pi iz/p}\). Under the hypotheses stated above, the authors show that \[ \max_{b\text{\;in\;}{\mathcal B}} | T(b)| =O_p(t^{3/4}p^{n/8}n^{1/4}) \] with an implied constant depending only on \(p\). The proof proceeds by splitting \(k\) in \(\mathcal B\) as \(k=u+vp^\nu\) where \(0\leq u<p^\nu\) and \(0\leq v<p^{n-\nu}\) so that \(\langle b,k\rangle =\langle d,u\rangle + \langle e,v\rangle\) where \(d=(b_0,\ldots,b_{\nu-1})\) and \(e=(b_\nu,\ldots,b_{n-1})\). This leads to \[ | T(b)| =\left| \sum_{u=0}^{p^\nu-1} e_p(\langle d,u\rangle) \sum_{v=0}^{V-1} e_p(s_{u+vp^\nu}+\langle e,v\rangle)\right| +O(p^\nu) \] where \(V=\lfloor tp^{-\nu}\rfloor\). After an application of Cauchy's inequality, the inner sum is again formed from a linear recurrence \((a_n)\) over \({\mathbb F}_p\) of degree \(n\) and period \(t\) and with an irreducible characteristic polynomial. This sum can be estimated by means of the Korobov bound: for \(1\leq U\leq t\) \[ \sum_{u=0}^U e_p(a_n) = O(p^{n/2}\log t). \] The case \(p=2\) has cryptographic significance since \(T(b)\) is essentially the non-linearity of the sequence \((s_k)\). However, the estimate in this case is not as strong as the best known bounds for the non-linearity of specially constructed sequences.
    0 references
    nonlinearity
    0 references
    linear recurrence sequences
    0 references
    Fourier coefficients
    0 references
    exponential sums
    0 references

    Identifiers