0
$\begingroup$

In the case where I am self-signing my digital certificates, if I just change my certificate (generate a new one) every week or month, will this have (sort of similar to) the same effect as implementing Forward Secrecy, since I should have a new private key every time?

Please correct me if I'm mistaken.

Thanks.

$\endgroup$
  • 1
    $\begingroup$ If the certificates use a new private key each time and you erase the old private key (tricky), you get a weak form of forward secrecy. If you create a new certificate with the same private key you get no forward secrecy at all. $\endgroup$ – CodesInChaos Mar 11 '15 at 9:08
1
$\begingroup$

This only works if you are absolutely certain that you're properly erasing the old private keys. This is harder than it seems; you have to be sure that you're leaving no traces anywhere, including on backup media or scattered around your hard drive. With proper forward secrecy methods, the keys never leave RAM, which makes erasing them securely much, much easier. Furthermore, this only happens after a week or a month or whatever; if you erase the private certificate monthly, then an attacker who acquires the certificate at the very end of the month (before you erase it) can decrypt the whole months' worth of messages.

There is very little reason to do this. ECDHE is not slow; while it's a bit slower and more computationally intensive than RSA, that difference shrinks rapidly as security increases (at 2048-bit RSA keys, there's a very small difference). Furthermore, ECDHE does not have anywhere near as many implementation pitfalls as securely erasing a private certificate weekly, and doesn't leave the week- or month-long interval for private key theft. Just use ECDHE.

$\endgroup$
  • $\begingroup$ You might want to clarify the second paragraph a bit. The diffie-hellman part of ECDHE is faster than RSA. It's only ECDHE_RSA which involves an RSA signature in addition to ECDHE that it gets slower than plain RSA. With ECDHE_ECDSA you should be even faster than plain RSA. $\endgroup$ – CodesInChaos Mar 12 '15 at 11:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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