Re: Dishonest Forks? | Was - Re: [Namedroppers-honest] Brian Smith asks: Who is Dean Anderson?

On Wed, 10 Mar 2010, Dean Anderson wrote:

> This paper expounds on "Kaminsky-class" cache poisoning.  As Kaminsky
> was discovered to be a hoax at the same time (Feb 2009) as this paper
> was published; the basis of the paper is likewise unsound. The paper
> you cite does not demonstrate any flaw in DJBDNS.  After that, it
> looks like a generic paper on cache poisoning.  Its no proof of what
> you assert.  The paper is just a sad example of people getting fooled
> by Kaminsky.

BTW, the paper does in fact mention the mention the King patch, but
provides no math to support its contentions. The paper does claim that
the King patch "removes the birthday window" and that claim can be
refuted on its face. One can never remove the possibility of a birthday
attack, one can only compute its chances of success. On that basis, the
paper, on the subject of DJBDNS, is incorrect.

I just sent to the list a couple emails sent to Kaminsky and Day, that
debunked their original claims back in October 2008.

The piggybacking scheme re-uses ports for multiple queries. This means
that several queries share the same port. The change reduces the number
random bits added to each query.  It means that attackers can
potentially discover the port with another query and that several query
responses can now match the same port.  While the math is harder to
evaluate in this case, it basically has the effect as though the
attacker gets more tries, since there are fewer combinations of ports
and queries. Each attack packet can potentially match several queries.
It also changes the birthday model math slightly.  The math is harder to
evaluate because the urns are now different sizes, and evaluating that
formula requires a complicated summation. 

But to get a feel of what happens, the replacement formula for n1 == n2
(urns equal) is: (this is a simpler formula than the non-replacement
formula, and is a lot easier to remember):

  1 - exp ( - n^2 / m )

For 200 ports: 
(%i9) float ( 1 - exp ( - ( 200 ^ 2 / ( 64510 )))); (%o9)
0.46208724940977 

For 400 ports
(%i10) float ( 1 - exp ( - ( 400 ^ 2 / ( 64510 ))));
(%o10) 0.91627650386183

The paper does mention that increasing the number of ports makes things
worse for a birthday attack.

Conclusion: the changes are worse.

		--Dean


-- 
Av8 Internet   Prepared to pay a premium for better service?
www.av8.net         faster, more reliable, better service
617 256 5494