Re: [tor-talk] Traffic shaping attack
On 3/18/16, Oskar Wendel <o.wendel@wp.pl> wrote:
Let's set up a service in a way that it will modulate the traffic, so the download would look like:
That's active manipulation in / at one endpoint node.
Then, we monitor traffic flowing into various entry nodes (remember we're a global adversary, having direct access to infrastructure around the globe) and spot the traffic that matches our pattern.
That's global / regional passive listening, needing be concerned minimally visibility with just any other G/R IP endpoints without needing track entire path. Which, if presumed and likely to be deployed, combine to be nicely effective, whether finding such clients, or services on Tor, I2P, etc. Attack could be made much more difficult quite possibly defeated if all nodes engaged in bucketed reclocked and jittered fill traffic with each other (possibly along some virtual path distance >=1 hop) and enforced peering relationships based upon receipt of same expected and contractually obligated traffic (would you talk to or retransmit for a node that acted sent packets as you say... fuck no).
Tor and other overlays also uses only TCP, which may provide some sort of enhanced end2end specific directable observability stream container. Opposed to UDP or raw packet which may offer more options to develop / layer various anti active / passive attack methods, including traffic spreading and recombination. Do not expect any possible solutions to come without potential overhead / tradeoff costs, only if you can use them to your needs, or adapt your needs to live in their context. And never underestimate a lorry full of tapes.
participants (1)
-
grarpamp