It doesn’t look that simple to me. From the Terrapin paper:
Although we suggest backward-compatible countermea-
sures to stop our attacks, we note that the security of the SSH
protocol would benefit from a redesign from scratch. This
redesign should be guided by all findings and insights from
both practical and theoretical security analysis, in a similar
manner as was done for TLS 1.3.
It seems the protocol itself needs a revision and implementation-specific patches are easier and less-than-ideal solutions.
One could argue that even these solutions they provide are already changes to the protocol, and not just fixes to implementation bugs. Both the Sequence Number Reset and Full Transcript Hash add or change functionality at the communication protocol level, rather than simply covering corner cases.
Yeaaaa, a complete redesign from scratch sounds way more dangerous. “Noah, get the boat” isn’t always the best answer. There’s been a lot of thought and testing put into the magnificent work that is SSH over the past few decades.
I won’t pretend I know better than the paper authors, what I can say is that some fixes are not incremental.
There are cases that mature tools and protocols should be left behind, and the danger lies exactly in using a protocol that was designed in the web 1.0 era.
Yes I was wrong to say that this an implementation detail rather than a protocol problem as the OpenSSH release notes to prevent this vulnerability include extensions to the SSH Transport Protocol, however I still believe that the headline is sensationalist at best since it can and has been protected against by patching ssh clients and servers. It would be entirely unreasonable in the majority of cases to simply stop using SSH on the basis of this vulnerability and that’s why I think the headline exaggerates the problem. The Register has a much more measured take on this including comments from the paper’s authors that people shouldn’t panic and try to fix immediately.
We went from “the fundamentals have changed” to “the 90s were a long time ago” real fast. Regardless of who made the point initially you are arguing it. Full redesigns are expensive, inefficient, and likely to introduce new vulnerabilities. The existing implementation is refined by decades of real world use. We can incorporate new lessons without a full redesign - if we can’t then we should stop being software engineers.
A full redesign is usually the type of project a CTO I worked for pejoratively called “computer science projects.”
TLS and SSH has quite different attack vectors so sure, basing SSH on TLS 1.3 would prevent the problems SSH has, but also bring in the problems TLS has. Thing is, I much prefer SSHs tradeof for things SSH is used for while TLS could be argued makes a lot more sense for the HTTPS use case. It just very different chains of trust with very different weak points, just pointing at TLS 1.3 as a solution when talking about SSH is quite ignorant.
It doesn’t look that simple to me. From the Terrapin paper:
It seems the protocol itself needs a revision and implementation-specific patches are easier and less-than-ideal solutions.
One could argue that even these solutions they provide are already changes to the protocol, and not just fixes to implementation bugs. Both the Sequence Number Reset and Full Transcript Hash add or change functionality at the communication protocol level, rather than simply covering corner cases.
Yeaaaa, a complete redesign from scratch sounds way more dangerous. “Noah, get the boat” isn’t always the best answer. There’s been a lot of thought and testing put into the magnificent work that is SSH over the past few decades.
I won’t pretend I know better than the paper authors, what I can say is that some fixes are not incremental.
There are cases that mature tools and protocols should be left behind, and the danger lies exactly in using a protocol that was designed in the web 1.0 era.
Yes I was wrong to say that this an implementation detail rather than a protocol problem as the OpenSSH release notes to prevent this vulnerability include extensions to the SSH Transport Protocol, however I still believe that the headline is sensationalist at best since it can and has been protected against by patching ssh clients and servers. It would be entirely unreasonable in the majority of cases to simply stop using SSH on the basis of this vulnerability and that’s why I think the headline exaggerates the problem. The Register has a much more measured take on this including comments from the paper’s authors that people shouldn’t panic and try to fix immediately.
Whole system rewrites are almost never a good idea
they are when fundamental assumptions change
Can’t expect rewrites to be automatically better than what we have now. We have so many replacement for Clang…
In what way have the fundamental assumptions of SSH changed?
SSH carries design choices from the 90s that might not apply today.
But it’s the paper authors themselves who are talking about a redesign, not a random Lemmy user, so idk.
Point is - a system redesign is very much something worth looking into if improving the existing system will be too disruptive.
We went from “the fundamentals have changed” to “the 90s were a long time ago” real fast. Regardless of who made the point initially you are arguing it. Full redesigns are expensive, inefficient, and likely to introduce new vulnerabilities. The existing implementation is refined by decades of real world use. We can incorporate new lessons without a full redesign - if we can’t then we should stop being software engineers.
A full redesign is usually the type of project a CTO I worked for pejoratively called “computer science projects.”
If you read the other article linked, there are literally already fixes available for many ssh implementations. Doesn’t seem that disruptive to me…
TLS and SSH has quite different attack vectors so sure, basing SSH on TLS 1.3 would prevent the problems SSH has, but also bring in the problems TLS has. Thing is, I much prefer SSHs tradeof for things SSH is used for while TLS could be argued makes a lot more sense for the HTTPS use case. It just very different chains of trust with very different weak points, just pointing at TLS 1.3 as a solution when talking about SSH is quite ignorant.