Ever since I read the Meridian research paper I’ve always been intrigued by overlay networks built using metrics that can’t be “faked” (latency in Meridian’s case).
One of the weakness of DHTs is that the distance metric is not tied to the physical domain. One category of attacks is feasible because one can “mine” the hashes used as identifiers so that they can position malicious nodes close to their target.
Meridian is not a DHT but can still be useful for gossip, discovery, etc… (advantage being lightweight and low latency).
I’m also thinking of possible mechanisms to prevent nodes from probing the whole network by limiting requests from nodes outside of a definite latency threshold. This would prevent peers from discovering the topology of the network.
I’m looking for possible attacks and weakness on this kind of overlay network (beside the reduced usefulness). I’m not a researcher but curious about what has been written on this topic.