RE: CRC32 - thoughts on Gutmann response

Eric M. Hopper wrote:
> Good security isn't just for you, it's for everybody you share
> the Internet with.  Once a tool can be shown to have weaknesses
> that can be fixed, they should be fixed, no excuses, no "it's
> good enough for my purpose", none of that.  That way leads one
> day to an Internet that's completely unusable for the vast
> majority of people because of DDOS attacks and whatever else.

Well, I didn't say it *shouldn't* be fixed.   But let's put this into
perspective here.

There's people that will tell you that 64-bit RSA just isn't enough
these days, because 128bit is "so much more" secure.  It is more secure,
but it took what, 5 years and hundreds of thousands of PC's, for
distributed.net to find their key.   I'd say 64bit is fine for me..

Padding the packets is fine, but not if it's going to give me 20K/sec
when now I get 30K/sec.  Changing the CRC32 to something else is fine,
but not if it's going to again get in the way of actually using the
tunnel, all in the name of a percieved security threat..  

I say percieved because this so-called security audit simply pointed out
possible weaknesses in the security of the tunnel, not an example of how
to exploit it.   As we stand right now, the only way to break the thing
is to brute force the key, and that's not feasable.  And you *can* brute
force any key.

I don't think many of us believed that a perfect cipe tunnel and a
perfect ipsec tunnel would be equal in security.  If you truly need a
100% secure system over the internet (well, close to it at least) then
you'd probably be using something else.   Or maybe not the Internet at
all.  I guess I just don't understand all the current hype about this.

It really all depends on what you need.  Consider the traffic that will
travel through the tunnel.  If it's really not more then what people are
sending completely unencrypted over the web or via SMTP, then all the
security doesn't even matter anyways... It might just be nice to have a
firewalled tunnel instead of using direct connections over the Internet
and NAT.   In it's current form, cipe performs quite well in these
situations and many more.

Again, I'll say again that I'm not saying that these fixes shouldn't be
implimented..

Ps.  I enjoy reading a good discussion.   Thanks!   Not sure I
understand the DDOS comparison though.

-----Original Message-----
From: Eric M. Hopper [mailto:[email protected]] 
Sent: Friday, September 26, 2003 2:27 AM
To: [email protected]
Subject: RE: CRC32 - thoughts on Gutmann response


On Fri, 2003-09-26 at 00:47, Joseph Jamieson wrote:
> To more people then you would think, security just isn't a major 
> factor. Cipe is very easy to install and run, it's solid as a rock, 
> and it includes some good encryption.  Not to mention it's nice and 
> fast.
> 
> It's not the perfect security model, but for a good many people, it 
> Just Works.
> 
> At my company, we have set up cipe between many small offices quite
> successfully.   They do e-mail and other such things over it, usually
> with low end DSL lines and cable modems.  Heavy duty unbreakable 
> security just isn't too important; if someone is going to spend enough

> time to crack the session and read the data, they'd see a whole lot of

> nothing.

Good security isn't just for you, it's for everybody you share the
Internet with.  Once a tool can be shown to have weaknesses that can be
fixed, they should be fixed, no excuses, no "it's good enough for my
purpose", none of that.  That way leads one day to an Internet that's
completely unusable for the vast majority of people because of DDOS
attacks and whatever else.

I'm not saying that if CIPE isn't fixed right now that it's a total
disaster.  I'm just saying that this kind of argument against fixing
something isn't at all valid, and not worth paying attention to.

> If someone could come up with a VPN that was as firewall friendly, 
> fast, and as easy to set up as cipe, but with all the latest and 
> greatest security measures, great!  If not, I'll still use it, at 
> least until all the theory is proven and someone actually cracks cipe.

This, I sort of agree with.  What should happen is that CIPE should be
fixed.  It's possible that OpenVPN and CIPE should be merged.  IMHO,
IPSEC should be completely abandon because it's too complex to implement
securely or even interoperably.

The utter simplicity (comparitively anyway) of configuring CIPE means it
should stay around.  Complexity is the enemy of security.  The minor
modifications that would make it more secure aren't very difficult to
implement, and don't add significantly to the complexity of the protocol
or the complexity of the implementation.  Unfortunately, they do mean
that a new, incompatible protocol needs to be created.

> Nobody said cipe was a replacement for SSL/SSH.  What an ignorant 
> bastard.

Simply because his tone was snotty, snide and incendiary doesn't mean
the points aren't valid.  It is clear that he didn't study CIPE as
carefully as he could've, but he wasn't being payed for a security
audit, was he?  When one gets things for free, like the attention of
someone who mostly knows what they're talking about, it's good to try to
pull out the useful things and ignore the not so useful things.  That's
the lubricant that greases the Open Source world.

Have fun (if at all possible),
--
There's an excellent C/C++/Python/Unix/Linux programmer with a wide
range of other experience and system admin skills who needs work.
Namely, me. http://www.omnifarious.org/~hopper/resume.html
-- Eric Hopper <[email protected]>


--
Message sent by the [email protected] mailing list.
Unsubscribe: mail [email protected], "unsubscribe cipe-l" in body
Other commands available with "help" in body to the same address.
CIPE info and list archive: <URL:http://sites.inka.de/~bigred/devel/cipe.html>