Security & Resilience
Selfish mining, and the assumptions it needs
Protocol referenceSource: Eyal & Sirer, Majority is not Enough, 2014No hash-rate, revenue or profitability figure is quoted here; those are market data with a vintage this page does not carry.
- Written by
- The Bitcoin Data Guide editorial team
- Reviewed by
- Protocol review desk
- Dates
- Published 2026-10-31 · Last reviewed 2026-10-31
Fact
The published model shows a withholding strategy can exceed a miner's proportional share under stated assumptions about propagation and miner behaviour.
Forecast
Whether any real operator could profit from the strategy depends on network conditions and coordination that the model assumes rather than measures.
The attack model
The honest strategy is to publish a block the moment it is found, so the rest of the network builds on it and the finder collects the reward. The selfish strategy withholds. A miner that finds a block keeps it private and starts mining the next one on top of it. The rest of the network, unaware, keeps mining the previous tip. If the selfish miner then finds a second block, it publishes both at once and now has a chain one block longer than the honest one, which the longest-chain rule makes the network adopt.
The gain comes from the blocks the honest network wastes. While the selfish miner's chain is hidden, every honest block found on the old tip is discarded when the longer private chain is revealed. Those wasted blocks are work the honest miners did and were not paid for, and the selfish miner's share of the total reward rises above its share of the hash rate. The published analysis by Eyal and Sirer showed that this can hold for a miner with less than half the hash rate, which is what made the result notable.
The assumptions the result rests on
The strategy is only profitable under conditions the model has to assume, and the assumptions are the honest part of the story. The first is propagation: the model assumes the selfish miner can learn of honest blocks and publish its own before the honest network can react, which depends on how quickly blocks travel and how the miner is connected. A miner with poor connectivity cannot win the race it depends on.
The second is the behaviour of other miners. The model assumes that when two chains of equal length are visible, a fraction of honest miners will build on the selfish miner's chain rather than the one they saw first. That fraction is a parameter, not a measurement, and the profitability threshold moves with it. If honest miners reliably build on the first block they see, the strategy's advantage shrinks.
The third is that the strategy is detectable and costly to sustain. A miner that repeatedly orphans honest blocks is visible in the orphan record, and the response available to the rest of the network is to reject its blocks or to coordinate against it. The model treats the miner as anonymous and the network as passive, which is a simplification rather than a description of a live network.
What the defence actually is
There is no consensus rule that forbids withholding a block, because a block that is never published is not a message the network can reject. The defence is economic and social rather than a line of code. Faster block propagation reduces the window in which a private chain can win a race, which is why relay improvements such as compact block relay matter to the security argument and not only to latency. A miner that is seen to orphan honest blocks repeatedly risks its blocks being treated as hostile.
The practical reading is that selfish mining is a real result about the incentive structure, and a weak attack in practice. It is worth understanding because it shows that the longest-chain rule is a coordination convention rather than a physical law, and that the convention holds because the participants keep choosing it. The 51% attack page covers the case where a miner does control a majority, and the miner cheating page sets out the boundary of what a block template can change.
Sources
- Eyal & Sirer (2014)
- Majority is not Enough: Bitcoin Mining is Vulnerable — the original selfish-mining analysis and its stated assumptions.
- Nakamoto (2008)
- Bitcoin: A Peer-to-Peer Electronic Cash System, section 5 — the longest-chain rule the strategy tries to exploit.
Related reading
- Security & ResilienceHow the network resists attack, and what happens when it is stressed.
- Can Miners Cheat?What a miner can and cannot do with the block it finds, and why invalid blocks fail.
- Double SpendingHow a double spend works, and why confirmation depth is the defence.
- 51% AttacksWhat a majority of hash rate can achieve, and what it cannot.
- Eclipse AttacksHow a node can be isolated from the honest network, and the defences against it.
- When Your Node Goes OfflineWhy a disconnected node is not a threat, and how it catches up on return.