Michael Richardson writes:
> 
> Tero Kivinen <[email protected]> wrote:
>     > 3) Client can also be smartphone, i.e. device which have quite a lot of
>     > CPU power and/or memory, but does not want to use it as it would
>     > increase the power usage so much that the battery life will be
>     > shortened.
> 
> Except that client being smartphone/etc. is behind NAT44, and along with
> ten thousand other smartphones, all have the same IP address... this matters
> because that scenario is probably indistinguishable from:

Good, as then we get rid of the NAT44... And then I woke up...

Anyways, I would expect big CGNs to try to spread the clients over
multiple different IP-addresses, instead of putting them all to single
IP-address. There are lots of other things that gets banned by
IP-address, so this kind of spreading is something they should do
anyways. For example forums, games etc quite often ban specific
IP-addresses for a while after certain bad events.

I know that slashdot once banned our company IP-block because some
people fetched their mainpage too often...

>     > 7) Botnets have huge amount of CPU power and lots of memory, but still
>     > limited number of distinguished IPv4-addresses or IPv6-prefixes (it
>     > might be millions, but most likely around thousands or tens of
>     > thousands IP-addresses).
> 
> a situation where there is an enterprise of compromised systems behind
> the enterprise firewall. (University lab networks come to mind...)

In which case it is good thing that people start complaining to the
helpdesk, which will start investigating the issue, and they find out
they have botnet machines inside the network...

I.e. it is almost impossible to protect against the attacks where the
attacker is inside your own machine or located very close to you...

> Further, the botnets don't need to present thousands of distinguished IPv4
> addresses, they can present a small number of attack nodes, spreading the
> work across the botnet?

You assume that the cost of puzzle would be meaningful for them so it
would be useful to spread the work. I assume it will not be. Most
likely the puzzles need to be something that the small devices can
also solve, which means they are not that expensive for the real
desktop machines.

On the other hand if sending a single packet costs 1 unit of CPU
power, and attacker has 10000 units of CPU power in his machine, that
means he can send 10,000 packets per second out. If the puzzle
requires 100 units of power, that means we reduced the attack speed
down to 100 packets per second. For the SGW verification will most
likely take something like 1 unit, so it can withstand attacks of 100
machines with same cpu power as it has.

If the small device have 10 units of CPU power per second, that means
it will take 10 seconds to solve the puzzle, but as I said the small
device will most likely be able to cope with such long delay for
connection, and it can still connect.

Spreading the attacks to 1000 other machines does not really help as
the network overhead to distribute the puzzles will get more expensive
than solving the actual puzzle...

Also the attacker will need lots of different IP-addresses as those it
has used before will get blacklisted. There is also ways of using
constant memory to store this kind of blacklists
(http://en.wikipedia.org/wiki/Bloom_filter) which are already used for
the graylisting and such mail servers to filter out requests by
IP-addresses. 

> So this tells me that we are looking for puzzles which are *not*
> parallelizable.   I know little about this kind of stuff...

If the puzzle is very expensive then yes, but if it just to raise the
relative cost of the attacker to higher, then they most likely will
not have reason to distribute it.

Also the amount of unique source IP-addresses is much bigger problems
to them than CPU power...

> I suspect, however, that the simplest machines to DDoS will be the
> smallest gateways.

Of course. On the other hand the attackers have less interest to
attack those. It does not really get to news when they tell that they
DDoSed some persons home gateway. Most likely even if they tell that
they DDoSed 10000 home gateways is not really news... Most of the
users would not even notice this, they would just blame their ISP or
vendor as the device is slow, and then they would reboot it :-)

I would expect the attack to go against big enterprices, for example
the power or phone companies.
-- 
[email protected]

_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec

Reply via email to