🧵1/ Filters can work iff they collectively form a vertex-cut (or many). Their goal is to partition the gossip flood into multiple components for some messages. It is thus wrong to think of it as a single node (independent) effort, it's a distributed attempt to form a vertex-cut.

Jul 13, 2026 · 1:10 PM UTC

2
5
9,719
2/ Collapsing each non-filter connected component to a single green vertex shows it forms an eclipse topology. If we get two or more non-filter components (or eclipses when collapsed) then we have (successful) message flood censorship between these components.
1
2
137
3/ Filters are a mechanism to suppress the gossip of certain messages at the p2p layer. This can be done for good reasons like protecting the nodes from DoS vectors or because we hold subjective opinions about which _valid_ transactions we don't like.
1
2
114
4/ The former class of filters completely stops working once you have a gossip algorithm that makes it hard to form vertex-cuts. We can't prevent nodes from trying to form it, but we can make it much harder by adding more connections between the nodes.
1
1
111
5/ TLDR; filters like the op_return relay limit and similar aren't really a solution that scales well over time. Not at 80b not at 100kb. It's more like a topological exploit that's really only possible to "use" while the network topology is weak against vertex-cut partitioning.
1
1
5
168
6/ More here. Please correct me if you find mistakes. claude.ai/public/artifacts/e…

Relay Filters as Eclipse Attacks: P2P Censorship Explained

Explore how Bitcoin relay filters form vertex cuts that eclipse honest nodes. Interactive graphs reveal the topology of p2p-layer censorship and its network effects.

claude.ai
23
Sort replies: Relevant Recent Liked
Replying to @phyrooo
Filters = censorship A squishy unlocked protocol = not decentralized btc ≠ Bitcoin
1
4
112
Load more