P2P (peer to peer) networks are everywhere. From software networks (bitTorrent, etc.) to distributed sensor networks (which are being installed everywhere) to human networks (from decentralized terrorist/guerrilla groups to online communities). One of the biggest threats to a P2P network is a bad node. A bad node -- which occurs through either malicious insertion or malfunction -- can block/capture/falsify information flow in the network and thereby threaten its very existence.
- Blackballing and Jury systems. Voting systems that identify and "vote off" bad nodes are very costly in terms of resources. Good nodes are forced to expend resources while bad nodes are not. Universal polling is required, etc. In the human sphere, this is the equivalent of getting a jury of peers to unanimously convict (that's morally tough).
- Buddy list systems. A voting system that promotes good nodes (other than yourself) to a trusted level is much less costly. A simple way to do that is to share buddy lists. As in: here are the people I like to work with. In order to be considered part of the core network, the bad node needs to work much harder.
- Suicide attacks. One of the most efficient ways to eliminate bad nodes is to allow other nodes to suicide them. In short, a good node recognizing bad behavior in another node can opt to disconnect themselves and the bad actor from the network. IF node A (a good node) recognizes that node B is bad, node A can opt to inform the network that both A and B are not operable anymore (essentially a suicide bomb). Given the limited range/scope/scale of attack, this is likely a very inefficient vector for attacks by bad nodes, so reversal of the process has limited downside.