Re: [tor-dev] Proposal Waterfilling

On Mar 7, 2018, at 5:12 PM, Florentin Rochet <florentin.rochet@xxxxxxxxxxxx> wrote:

Hello,

On 2018-03-07 14:31, Aaron Johnson wrote:
Hello friends,

1) The cost of IPs vs. bandwidth is definitely a function of market
offers. Your $500/Gbps/month seems quite expensive compared to what
can be found on OVH (which is hosting a large number of relays): they
ask ~3 euros/IP/month, including unlimited 100 Mbps traffic. If we
assume that wgg = 2/3 and a water level at 10Mbps, this means that,
if you want to have 1Gbps of guard bandwidth,
- the current Tor mechanisms would cost you 3 * 10 * 3/2 = 45 euros/month
- the waterfilling mechanism would cost you 3 * 100 = 300 euros/month

The question of what the cheapest attack is can indeed be estimated by
looking at market prices for the required resources. Your cost
estimate of 3.72 USD/Gbps/month for bandwidth seems off by two orders
of magnitude.

Let me merge your second answer here:

I see that I misread your cost calculation, and that you estimated $37.20/Gbps/month instead of $3.72/Gbps/month. This still seems low by an order of magnitude. Thus, my argument stands: waterfilling would appear to decrease the cost to an adversary of getting guard probability compared to Tor’s current weighting scheme.

There is still something wrong.

What’s wrong? $37.20Gbps/month = 30 Euros/Gbps/month, which is what you are claiming. This would be the lowest price for a sustained Gbps transfer by a significant margin among all of the deals that have appeared on this thread. The other lowest was from Alex, who found $100/Gbps/month. I somewhat doubt that you could actually achieve 1Gbps sustained for 30 Euros/month on a shared VPS or that OVH would actually tolerate using this much bandwidth at this little cost. It would at least be a notable new record for the cheapest possible Tor bandwidth, as far as I can tell.

With Waterfilling, we assume above a water level of 10 Mbits, so we need:

100 VPS SSD 1 relaying 1Gbps at the guard position, which the cost turns
to be 3*100 = 300 euros/month.

This calculation is much too kind to waterfilling :-) Why pay for a full 100Mbps with only 1 IPv4 address when you only need 10Mbps/IP to achieve the same guard probability? Earlier I showed an example of a cheaper VPS (https://my.hiformance.com/cart.php?a=add&pid=165) that appears to provide for just $0.63/month a VPS with an IPv4 address that is capped at 6Mbps sustained througput. This would be a more economical way (3.5x cheaper) to attack waterfilling. Alternatively, I bet you could get bulk IPv4 addresses for much cheaper than the $3/month that OVH charges for its SSD VPS, which you could then potentially apply to your 100Mbps (or larger) server to get 10Mbps and more cheaply attack waterfilling. For example, OVH provides 256 IP addresses for its dedicated servers at no monthly cost (https://www.ovh.co.uk/dedicated_servers/details-servers-range-GAME-id-MC-64-OC.xml). These servers can be had for at least 55 euros/month, which provides 500Mbps unlimited. With two of those, you could achieve the above attack on waterfilling for 110 euros = $136.36/month instead of 300 euros/month = $371.92/month. Once we’re talking about trying to achieve a large fraction of the Tor network, which requires many Gbps in vanilla Tor, the fixed cost of a server becomes a smaller fraction of the total cost and the savings from the free extra IPs become greater.

That depends on the kind of policy that the Tor network could put in
place. If we decide that large families become a threat in
end-positions, we may just aggregate all the bandwidth of the family,
and apply Waterfilling. That would not kick them off, but would create a
kind of 'quarantine'. Same kind of suggestion than the one just below.

This seems to be a different argument than you were making, which was that the many servers must appear to be run independently, which I still disagree with.

This is what Waterfilling does: increase the cost of a well-defined
attacker and offer clients to choose into a more "diverse" network.

Sorry, I still don’t agree. It increases the cost in terms of number of IP addresses required and causes clients to spread out more across guards with different IP addresses. This is a narrow notion of diversity and not one that I think is useful as a design principle.

Best,

Aaron