Random Polynomials in Legendre Symbol Sequences

It is important in cryptographic applications that the “key” used should be generated from a random seed. Thus, if the Legendre symbol sequence generated by a polynomial (as proposed by Hoffstein and Lieman) is used, that is { (f(1)p),(f(2)p),(f(3)p),⋯,(f(p)p) }, \left\{ {\left( {{{f\left( 1 \right)} \over p}} \right),\left( {{{f\left( 2 \right)} \over p}} \right),\left( {{{f\left( 3 \right)} \over p}} \right), \cdots ,\left( {{{f\left( p \right)} \over p}} \right)} \right\}, then it is important to choose the polynomial f “almost” at random. Goubin, Mauduit, and Sárközy presented some not very restrictive conditions on the polynomial f, but these conditions may not be satisfied if we choose a “truly” random polynomial. However, how can it be guaranteed that the pseudorandom measures of the sequence should be small for almost "random" polynomials? These semirandom polynomials will be constructed with as few modifications as necessary from a truly random polynomial.