[messaging] Once again: Tor timing attacks and a Tor confession

[grarpamp]

On 2/29/16, Georgi Guninski <guninski at guninski.com> wrote:
> Replace "sufficiently decent" by "perfect", or define it to be "provably
> intractable" and do not assume hardness not proved unconditionally, like
> P != NP.

So long as each node accounts for negotiated contract rate
with peers, and generate fill for missing packets on the inbound
links when output the other side, and reclock all the input when
output to a fixed rate, and add random jitter to the output links
to mask time spend negotiating and compensating for the input
junk received... it would seem range from reasonable sufficient
to damn hard.

It's an enhanced level of the fixed bucket clocks in old school
ATM / TDM that people seem to forgot about...

https://en.wikipedia.org/wiki/Asynchronous_Transfer_Mode#Traffic_policing
https://en.wikipedia.org/wiki/Time-division_multiplexing

There was even talk on one of these lists about doing fill
not just in the overlay networks, but also doing it, along
with automatic pfs style encryption in the layer zero link
hardware itself (ethernet PHY, etc) by starting an IEEE / IETF
working group... every switch, router and NIC port everywhere.

Some OP threads for ref:
https://cpunks.org/pipermail/cypherpunks/2016-February/012436.html
metzdowd: "traffic analysis" Jan 2015
My spam on @cpunks @torproject
Etc et al

Encrypted fill traffic is at least worth thinking about, thus cc.