Thank you for the technical post. Hope this answers some.
Thanks for your productive response. I actually own both DRK and CLOAK, so want to understand the systems better.
Obfuscation. You have to think about PoSA like tor.
1. We can black list malicious nodes and rely on the fact that there will be so many nodes on the network (unlike centralized masternodes) that the majority will not be spy nodes and to find any relevant data will be next to impossible. Nothing is 100% approaching limits is the best way to handle things.
You can only black list nodes if you can detect that they are doing something bad. There is no way to detect which nodes are the malicious nodes that are spying because other than recording transactions, they perform their duties perfectly and are upstanding citizens according to the protocol. However, they secretly record the links between transactions and publish them publicly elsewhere (example, their own de-anonomised block explorer).
2. Same question as 1
Indeed.
3. 2 levels out of a random selection. powers of two magnify probabilities of not being interfered with exponentially.
The problem is that 2^1 is only 2. To leverage the power of two, you need to add more layers of anonymity. For example, 8 or 10 steps, so you get 2^10 = 1024. However, the problem is that would slow down transactions too much.
4. Again the same questions
No. It's quite different. My 4th question is about someone stealing funds, not anonymity.
Blacklisting is only a first defense and I have no idea if it is even going to be implemented, just a thought and it is a weak one.
About layers of anon.. you forget that you have as many nodes as you have on the network to randomly be assigned to.
We have already been thinking about ways of prevent nodes from knowing when a posa tx is incoming. Packet sniffing is an issue we are combating.