I have read about Feistel network and so far only advantages have been listed. Are there any disadvantages to the design? Please could you explain in fairly basic terms?
$\begingroup$
$\endgroup$
-
4$\begingroup$ "... so far only advantages have been listed" - where? This isn't wikipedia, but...citation needed. $\endgroup$ – Maarten Bodewes♦ Feb 18 '17 at 21:43
$\begingroup$
$\endgroup$
Well, what are we comparing Feistel networks to? The other major design paradigm is SPN: https://en.wikipedia.org/wiki/Substitution-permutation_network. In comparison to SPN, the cons are:
- Not a direct implementation of confusion and diffusion. Feistel networks have more structure that has no clear justification. (Some point to the security proof but if you assume a cryptographically secure round function, how wrong can things go? A better question is: "if your round function is weak, what does the Feistel network add?").
- Parallelism is about half of an equivalent SPN, which is a disadvantage for hardware implementations.
-
$\begingroup$ Can one say, that those are also the reasons why Feistel networks are hardly used any more these days? $\endgroup$ – mat May 12 '17 at 10:00
$\begingroup$
$\endgroup$
I have found a disadvantage because at least 3 rounds of the Feistel cipher are required. More details can be found here: Is this a structural weakness of Feistel networks? - the computation cost will be higher.
-
1$\begingroup$ This is close to a link-only answer to the same site. You should flesh it out a bit: compared to what is this a disadvantage? $\endgroup$ – Maarten Bodewes♦ Feb 18 '17 at 21:44
-
2$\begingroup$ generally all ciphers require multiple rounds, and usually quite a bit more than 3, I do not consider that a disadvantage unless there is no Feistel cipher that is as secure as some other cipher in the same range of performance $\endgroup$ – Richie Frame Feb 19 '17 at 7:03
-
$\begingroup$ How is this a disadvantage? Three rounds is not very much. $\endgroup$ – forest Jun 19 '19 at 3:32
