3 ms·
This flaw doesn't hurt your service, and you do more with TLS than most people here. I do? The only thing I use TLS for is the Tarsnap website. The odds of y
by cperciva 17y ago
This flaw doesn't hurt your service, and you do more with TLS than most people here.
I do? The only thing I use TLS for is the Tarsnap website.
The odds of you or me or anyone else on Hacker News developing a crypto transport for which an attack as slick as this would be required are basically nil.
Tarsnap's transport layer is described and implemented in the public Tarsnap source code. If you're so sure that there are easier attacks against it, please go find one. :-)
This isn't so much a TLS problem as it is an HTTP + TLS problem.
Agreed. The whole thing is a layering violation, and it's not surprising at all that it's screwed up.
name the widely-used cryptographic transport protocol that someone invented that didn't subsequently have very serious flaws uncovered in it.
As JoachimSchipper points out, the SSH transport protocol has done rather better than SSL. More importantly, however, you're unreasonably limiting things by restricting consideration to widely-used protocols: In general, widely-used protocols tend to be protocols designed by committee which contain everything except the kitchen sink -- and it's that precisely that excessively large feature set which makes SSL so hard to get right.
- tptacek 17y agoSSH was gameovered in v1, and the current protocol has had serious findings. I could argue that the worst SSH flaw is worse than the worst SSL flaw, but I don't have to; that's not my point. Instead of answering my question, you editorialized. I'll assume you can't name a widely-used crypto-secured protocol that hasn't had a serious flaw discovered. That's OK, because I probably can't either. It took over 10 years for anyone to find this most recent flaw in HTTPS. You, for instance, didn't know about it. That's OK; neither did most cryptographers and protocol designers. What's weird is that from this fact you manage to draw the conclusion that people should just design their own protocols, as if they would do better without 10-15 years of attention from hundreds of academic and industry researchers. I happily recomment Tarsnap to people, Colin, but to my mind the reason there are no findings in the Tarsnap protocol is that a finding in Tarsnap is worth $0 to an industry researcher and 0 cites to an academic researcher.
- cperciva 17y agoYou, for instance, didn't know about it. Quite true -- but my attention rather drifted once I got past page 50 of the RFC. I would argue that any cryptographic protocol which cannot be described in less than 10 pages is too complex to be secure. a finding in Tarsnap is worth $0 to an industry researcher Come on Thomas, I'm sure it would be worth at least $1 to you to have me shut up about SSL. :-)
- tptacek 17y agoThat would be worth much more than $1 to me, personally, but unfortunately not worth enough to offset my bill rate. Your first argument --- crypto protocols that take more than 10 pages are insecure --- is, I think, asinine. Not you, just the argument. The argument is asinine. You're great. Argument: not so much. Why? Because more damage has been done to Internet security by crypto protocols that can be described in under 10 pages than has ever been done by TLS. Start with PPTP, RADIUS, and RDP, but I can keep going.
- cperciva 17y agoI didn't say that protocols which can be described in under 10 pages are necessarily secure! However, if you can describe something in less than 10 pages, it's probably going to be obviously insecure, rather than having problems lurking unnoticed for a decade.