3
$\begingroup$

I have been studying Cryptographic Boolean Functions and got stuck with the calculation of cross correlation of two boolean functions. I know the formula is $$c_{f,g}(y) = \sum_{x \in V_{n}} f(x)\cdot g(x\oplus y).$$ What I don't understand is if $f = g$ and $y$ is $0$ shouldn't the correlation be $2^n$? When I calculate, if $f$ and $h$ has the value $0$, $f(x)\cdot g(x)$ will yield $0$, but shouldn't I have $1$ in order to reach $2^n$ in the end? I am sorry if this question sounds silly but I have really spent so much time trying to figure out. Any help would be greatly appreciated.

$\endgroup$
3
  • 1
    $\begingroup$ Where did you find this formula? I think $\sum_{x\in V_n} (-1)^{f(x) \oplus g(x \oplus y)}$ is more common and then it nicely adds up to $2^n$ and you can actually have negative correlation. $\endgroup$
    – Bla Blaat
    Dec 17, 2017 at 9:41
  • $\begingroup$ indeed, correlation should be able to be negative $\endgroup$ Jan 17, 2018 at 4:36
  • $\begingroup$ The answer is good, see the reference. $\endgroup$
    – kodlu
    Feb 17, 2018 at 0:30

1 Answer 1

1
$\begingroup$

In theoretical computer science literature, boolean functions are defined as $$ f\colon \operatorname{GF}(2)^n \to \{\pm 1\}, $$ and the usual link with functions defined as $$ \tilde{f}\colon \operatorname{GF}(2)^n \to \{0,1\} $$ is via $$f(x)=(-1)^{\tilde{f}(x)}.$$

Edit Here's a standard reference from the Boolean function literature.

scanpages

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.