problem with lightning security model

gistfile1.txt

ah hello
digitsu
cool, yeah, what you described was pretty much my understanding.  But I have a question.
digitsu
what if Dave discloses R, and then Carol becomes unresponsive?
digitsu
at this point, carol should be using R to pull funds from Bob, but she dies instead.
digitsu
then what incentive would Dave have to give back the money to Alice (through a return path) he is paid.
digitsu
Carol ends up being the payer.
digitsu
this means it would be a good point to DDoS carol once you disclose the R to her.
jl2012
Dave should not disclose R until he gets the hash locked contact from Carol
digitsu
sure.
digitsu
and he gets that. which is why he is paid.
digitsu
but then Carol is DDoSed after that, so she cannot pass R to Bob.
digitsu
Bob reclaims his HTLC in 1 day.
digitsu
Carol ended up paying FOR Alice.
jl2012
That's another question. I don't have a good answer, other than "1 day should be enough". You may ask it on the lightning-dev channel
jl2012
Oh no
jl2012
If Carol can't communicate with Bob, she should settle the channel on chain
jl2012
Unless the bitcoin network itself is ddosed, or blocks are all full
digitsu
well, in a DDoS situalion carol cannot communicate with anyone.
digitsu
the solution in the paper says you need to trust a 3rd party to broadcast the closing txns for you.
digitsu
which works, but it somewhat centralized.
digitsu
and normally the 3rd party solution works as they are monitoring for old stale commitment txns (and will transmit the penalty txn if they see one)  But in this case, they have nothing to watch on the blockchain.  The fact that Carol is DDoS'd is silent.
digitsu
this attack vector is new and inherant to LN
digitsu
I asked this on Lightning-dev channel is slack.  Nobody answers.