for other stuff...do you have to ask? What sort of system do you think should be used for coordinating 'criminal' activity, instead of streaming super full SHD video for retards?
And the answer is : some sort of 'high latency' mixing network. And interestingly enough such a network doesn't seem to exist, although it seems to me it would require less resources than something like tor. And nobody seems to be worried about having or not having that kind of network, which strikes me as odd...
What I was trying to say is that, if the use case is 'criminal activity', then using a 'low latency' network like tor which provides centralized 'hidden' services is a not a good idea. It's more like a recipe for disaster.
https://www.aclu.org/files/natsec/nsa/Tor%20Stinks.pdf If the mix or other network design is doing it's job, programmed delay, being a part of observed latency, might not end up mattering much to security. ie: Think of latency as the clock speed of the network. The job of passing around cells from here to there is a compute task, a fixed todo list set into instructions. Doesn't matter whether your job completes in 10 seconds, or 10 days, or with each instruction artificially delayed by a billion nop's on the fast cpu. Your cpu can still be probed and watched for months regardless how long it takes, so it doesn't seem that waiting n time units before passing a cell from a to b, b to c, c to d, will help in itself, because the network, or your path over it is not guaranteed to be full. Enough of those odds of empty voids happening at the right points, the ones your data is moving over, and you're hosed. If you've got all this idle time, might as well fill the mix, an actual defensive move. Zen's remindful blurb something others suggested for longer, me included. It really needs cross checked against available research, and researched into development if it works.. Would also like to see work on opensource IETF standard for all physical ports on all network hardware, each port having independant full time random keyed link encryption with tamper alerts, and full time fill to the capacity of the link with random data, baked in the silicon and silently on by default out of the box. Limited to opposing crypto suites, say three with a q resistant one. The silicon, firmware, and software required is expected to be trivial cost, like $1/port when applied globally. That'll seriously fuck with the vampires on the wire who don't have legal authority to force you to turn it off, or to particularly regulate your hardware, which in the western world is many jurisdictions. To wit, apple secure enclave is still thought by many to be resistant and live years later in T2 and A11. No, operators don't need to turn it off to debug their net, they copy the port above the MAC PHY of the chaff at the logical layer to another port. Which also is a human action which can't currently be forced, in any sane regime, and fishing for win with an illegal general warrant.