Hey all, I just read through the whitepaper and have some questions about a 'split attack' - which seems easiest to execute to me in that you just need to issue double-spend transactions simultaneously with a tip-selection alg that hits different areas of the tangle (am I understanding that properly?). Is sounds like the resolution to that problem is one of the (large) splits becoming orphaned: does this mean that all transactions in that split chain are rejected? If somebody spammed split attacks, would every other transaction be rejected alongside it?
you have to keep both branches get going in order to make a double-spend. Because the weights of branches give unequal probability to be selected (or abandoned), the majority of honest nodes will go along with one with the heaviest weight and it would be harder for you to keep the other branches unless you own 1/3 of the network nodes. because the network always has latency among physical locations of the nodes, some transactions might occasionally choose other transactions from the orphaned branch but they will not be confirmed as transactions start selecting one another from the main branch immediately.
this is as far as i understood. someone corrects me if im wrong.
Thanks for the response tawago! I understand how the problem of double-spending occurring is not a huge concern, but I am more asking about the transactions that just happened to attach to a now-orphaned branch - are they rejected? Sorry if your post addressed this and I just misunderstood
"rejected" is not a correct term, i think. they will be still attached to the tangle but will not be referenced from other transactions, thus, never get "confirmed"