>> The document claims this is because "IPSec is too complicated to be 
>> secure", and that "TSL is mature and battle-tested".
> 
> I like how they make this assertion, then later on say "you might need 
> the load balancing that IPsec does, but you can get that with OpenVPN by 
> running this other complicated program on a spare machine." It sounds 
> like a lot of the complication is stuff that OpenVPN basically leaves out.

Well, is that a protocol feature or a software feature?

> Plus, I'm not really sure how they're running TLS over UDP, given that 
> TLS is stream-oriented and assumes reliable delivery. It's also not real 
> obvious from their descriptions that it's possible to run a UDP protocol 
> over OpenVPN.

Hmm, that's a good point.

>> Also amusing is the statement "Blowfish is a very strong algorithm 
>> with no known weaknesses. Its 128-bit key provides us with a large 
>> enough key space to make brute force key attacks impossible in 
>> polynomial time." Erm... like... WTF?
> 
> Of course, it hasn't been tested as furiously as AES, either.

I was more amused by the statement that key size has any relationship to 
complexity class.

Blowfish is far more popular than, say, TEA or SQUARE or any number of 
other ciphers from the zoo of less-known algorithms out there. I note 
however that Blowfish has been "replaced" by Twofish which is meant to 
be stronger. (And AES finalist, I believe.)

A lot of people are apparently jumpy about the whole XSL attack thing on 
AES.


-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
>>> The document claims this is because "IPSec is too complicated to be 
>>> secure", and that "TSL is mature and battle-tested".
>>
>> I like how they make this assertion, then later on say "you might need 
>> the load balancing that IPsec does, but you can get that with OpenVPN 
>> by running this other complicated program on a spare machine." It 
>> sounds like a lot of the complication is stuff that OpenVPN basically 
>> leaves out.
> 
> Well, is that a protocol feature or a software feature?

I'm saying that they mock IPsec for solving problems *they* don't have, 
because it was "designed by a committee".  I suspect the people on the 
"committee" actually *did* have those problems, so incorporated 
solutions to them into the standard, instead of making everyone solve it 
themselves with ad hoc solutions consisting of interacting layers which 
could easily introduce a security hole if you don't know what you're doing.

> I was more amused by the statement that key size has any relationship to 
> complexity class.

Yes, I understood the WTF. :-) Of course, if it's running over TLS, you 
can put whatever cipher both sides agree on in there.

And actually, key size *can* have a relationship to complexity class, 
perhaps. If you can pre-compute something that lets you look up in 
polynomial time the key that someone is using, except that the key is 
too long to store the precomputed somethings for every key, I can see 
where that can happen. (Technically, the same complexity class, but in 
practice, you can break something in polynomial time if you discount the 
precomputation, perhaps.)

But yeah, the whole discussion was full of WTFs like that. Plus, they 
say basically "IPSec is too complicated to deploy, OpenVPN does the same 
thing only its easy to deploy", without giving any evidence at all of 
either. Most of the paper is a description of SSL, and I'm pretty sure 
there aren't fundamental security problems with IPSec that aren't in SSL.

(And apparently the authors don't know that actual difference between 
SSL and TLS.  Hint: OpenVPN seems to use SSL.)

-- 
Darren New / San Diego, CA, USA (PST)
  Helpful housekeeping hints:
   Check your feather pillows for holes
    before putting them in the washing machine.

Post a reply to this message

Darren New wrote:

> But yeah, the whole discussion was full of WTFs like that. Plus, they 
> say basically "IPSec is too complicated to deploy, OpenVPN does the same 
> thing only its easy to deploy", without giving any evidence at all of 
> either. Most of the paper is a description of SSL, and I'm pretty sure 
> there aren't fundamental security problems with IPSec that aren't in SSL.
> 
> (And apparently the authors don't know that actual difference between 
> SSL and TLS.  Hint: OpenVPN seems to use SSL.)

Hmm. I just wanted to know whether this piece of software was worth using.

I think I have my answer now...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
> Hmm. I just wanted to know whether this piece of software was worth using.
> I think I have my answer now...

Oh. I imagine if it solves the problem you have, it's probably worth 
using, assuming you can get it working and it's sufficiently widespread 
you trust it.

That the people writing the *paper* didn't write it in a convincing way 
doesn't really say much about the *code*.

-- 
Darren New / San Diego, CA, USA (PST)
  Helpful housekeeping hints:
   Check your feather pillows for holes
    before putting them in the washing machine.

Post a reply to this message

Invisible wrote:
> Weeeee....
> 
> http://eprint.iacr.org/2008/166.pdf
> 
> There's nothing like cryptanalysis to make you paranoid! ;-)
> 
It lead to a discussion in parliament here and a bad time for one of our 
ministers. Introduction of the country-wide OV chipcard (OV is the 
abbreviation for public transport) is postponed now.

At the last election in some cities voting was done again by paper and 
red pencil because somebody showed that you could listen to what people 
voted by putting a radio receiver close to it and a few more 
insecurities of the digital voting machines.

What these examples have in common is that they were badly designed and 
they relied on people not knowing the details as an important security 
measure. The latter is of course an absolute sin in cryptography, but 
apparently not (yet) when designing voting equipment or identification 
cards.

Post a reply to this message

andrel wrote:

> What these examples have in common is that they were badly designed and 
> they relied on people not knowing the details as an important security 
> measure. The latter is of course an absolute sin in cryptography, but 
> apparently not (yet) when designing voting equipment or identification 
> cards.

Security through obscurity = hmm, it doesn't work. ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Darren New wrote:

> That the people writing the *paper* didn't write it in a convincing way 
> doesn't really say much about the *code*.

True. But given that it seems to be a one man project... Well I'm sure 
it *works*, but that's not the same as being secure. [Not that any sane 
person is going to bother trying to hack my systems unless it's 
sufficiently easy.]

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
> True. But given that it seems to be a one man project... 

I see. I hadn't followed it up.  No, if the author of the code is the 
author of the paper, I'd definitely go with IPSEC. :-)

-- 
Darren New / San Diego, CA, USA (PST)
  Helpful housekeeping hints:
   Check your feather pillows for holes
    before putting them in the washing machine.

Post a reply to this message

>> True. But given that it seems to be a one man project... 
> 
> I see. I hadn't followed it up.  No, if the author of the code is the 
> author of the paper, I'd definitely go with IPSEC. :-)

It appears the report was written by somebody other than the author. But 
then, comments like "OpenVPN seems to be using SSL" make me suspicious. 
As does the fact that SSL and TLS both run over TCP, yet OpenVPN seems 
to be using UPD. That looks like a big enough deviation from the 
standard that I'm left wondering how many design decisions have been 
made with potential security implications.

It's not like OpenVPN just calls the OpenSSH crypto library to do its 
bidding; it looks like a complete new implementation.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On Fri, 04 Jul 2008 09:08:07 +0100, Invisible wrote:

> It appears the report was written by somebody other than the author.

That sentence made me laugh.  Of course, it's 3 AM here (almost 3:30), so 
I'm a little sleepy, but can I just say THANK YOU?  I *really* needed a 
laugh.

Tough night, and this really helped.

Thank you.

Jim

Post a reply to this message