+ require some proof-of-work as part of the client-to-client connection process (helps prevent 'Sybil' attacks).
Isn't that a brilliant idea? Like hashcash?
You would be required to hash the string of the transaction, with a proof of work, that would say, take 5 seconds to calculate on a modern PC. Checking the POW just like in bitcoin would be easy and very quick for the receiving machines, but would stop a flood attack of random data without the attacker having limitless CPU power.
I was actually thinking of a minute or three of proof-of-work on initial connection, not when submitting a transaction, but requiring some proof-of-work for every transaction submitted into the network IS a very interesting idea! Should be straightforward to implement, too (add a nonce and either a full or partial hash to the transaction)...