Closed drwetter closed 6 years ago
You did nothing wrong. We disabled lucky13 at the moment since we were not happy with our timing measurements. We are planning on readding this feature though. You can use TLS-Attacker 1.2 if you want the old version which supports lucky13
thx for you response.
If that'll take longer how about returning a different message to the user?
Yes i guess we should do that or atleast remove that part from the documentation
Hi @ic0ns ,
any update on this? The thing which would be the best of course :-) is Lucky13 would be enabled again.
If the finding is not certain I would be happy with Vulnerable:Uncertain
and e.g. if you can exclude it Vulnerable:false
.
Cheers, Dirk
ẁell the lucky13 attack is a timing issue which is related to the cbc mode. You can exclude a vulnerability if the Server does not support cbc at all. Otherwise it will be very hard to determine if a remote server is vulnerable or not. If we readd this feature it will probably only work "reliably" against local servers.
I hope i can find a student to implement this, but dont expect this feature in the next 1 - 2 months...
Hi Robert,
I wouldn't have asked in the first place if the remote side has no CBC ciphers. There's an awkward TLS stack I am currently examining. It appears kind of ancient and outdated. It offering CBC ciphers only, moreover it's a commonly used service.
I have interpreted TLS-Attacker as a tool which can detect a Lucky13 vulnerability remotely, for all other means I can use another project (testssl.sh) I started.
Your comment gave me the reason to go back and re-read the original publication from Royal Holloway. I realized that in fact one needs more more request for recovery at least than I remembered. There was still the question to me how many request is needed for a detection only. As you indicated the paper says it's difficult and a good network connection would be needed.
So, thanks for the info, Dirk
Yes, the attack is only hardly detectable in local environments since the differences are very low. The tests I ran previously were on the localhost.
The next version updates the wiki to indicate that we disabled this feature
What did I do wrong?