Picture this: you've just migrated your staking pool to new hardware. Both machines come online. Both see the same slot assignment. Both sign a block. You've committed equivocation, signing two distinct blocks for the same slot, and somewhere on the network a validator client has already noticed, packaged the evidence, and queued the punishment. You don't get a warning.
Proposer slashing is one of Ethereum's most precisely engineered penalties. Understanding it means understanding both what equivocation is and why Gasper treats it as a capital-level offense in economic terms.
One Slot, One Block, No Exceptions
Gasper, the consensus protocol underlying Ethereum's proof-of-stake layer, combines two sub-protocols: LMD-GHOST (which selects the canonical chain by following the fork with the heaviest accumulated attestations) and Casper FFG (which handles finality). Every 12-second slot is assigned exactly one validator as its proposer. That proposer packages transactions, signs a block, and broadcasts it.
Equivocation happens when a validator signs two distinct blocks for the same slot number. It can occur across forks: the validator's software sees a short-lived network partition, produces a block on one chain, then produces a competing block on another before the two sides reconnect. The key signs both. That double signature is the crime.
The two blocks don't even need to be malicious in the obvious sense. A redundant setup meant to increase uptime can produce equivocation entirely by accident. Gasper doesn't distinguish intent. The signature is the evidence, full stop.
How the Slashing Actually Triggers
Any validator on the network can submit a `ProposerSlashing` object to the beacon chain. This object contains both signed block headers: the slot number, the offending validator's index, and the two distinct signatures over two distinct block roots at the same slot. The beacon chain state transition function verifies three things before executing the slash.
- Both headers reference the same slot.
- Both headers are signed by the same validator.
- The two block roots are different.
All three must hold. Then the slash executes immediately in the next processed block.
The economic hit arrives in two parts. First, the slashed validator loses a base penalty equal to 1/32 of its effective balance, taken immediately. Second, a correlation penalty lands roughly 36 days after the slashing epoch, and this is where the protocol gets genuinely clever: that delayed penalty scales with the total ETH slashed across all validators in a window around the same period. Slash one validator in isolation and the delayed penalty is modest. Slash a coordinated group of thousands simultaneously and it can consume nearly the entire effective balance. The mechanism is deliberately lenient toward accidents and brutal toward coordinated attacks, which is exactly the right design priority.
The slashed validator is also forcibly exited from the active set and barred from withdrawing funds for roughly four months, ensuring the correlation penalty has time to land before anyone walks away.
A Worked Scenario: Two Operators, One Key
Consider two operators, Priya and Daniel, running validators for the same staking pool. During a system migration, the pool briefly runs the same validator key on two separate machines. Both are online, both receive the slot assignment, and both produce a valid signed block for slot 8,400,000. The two blocks carry different parent roots because each machine's view of the chain tip was fractionally different at signing time.
A third validator, running standard client software, sees both headers propagate across the gossip network. Its client automatically packages them into a `ProposerSlashing` message and includes it in the next block proposal. The pool's validator is slashed. The initial 1/32 penalty fires within seconds. Because no other validators were slashed in that correlation window, the delayed penalty is modest, perhaps a few percent of the 32 ETH stake. The validator is ejected from the active set and waits out the exit and withdrawal delay before recovering any remaining funds.
Total loss: small in absolute terms, painful in percentage terms, entirely avoidable. The protocol's lesson is not subtle: never run the same validator key on two machines simultaneously.
The Fork Choice Wrinkle Worth Understanding
Slashing and fork choice operate on different timescales, and that gap is where equivocation across forks gets interesting.
When a proposer equivocates across two forks, both blocks might briefly attract attestations from validators who hadn't yet seen the competing block. LMD-GHOST eventually converges on one chain as attestations accumulate. But here's what doesn't happen: the protocol doesn't automatically detect equivocation during fork choice resolution.
Fork choice and slashing are separate mechanisms. Fork choice picks a winner; slashing punishes the proposer afterward, once someone submits the evidence. A successfully slashed validator may have had their block included in the canonical chain. The slash isn't a retroactive erasure of the block. It's a financial penalty on the key holder.
This separation is intentional, and I'd argue it's one of the more underappreciated design decisions in Gasper. Conflating fork choice with punishment logic would make the consensus mechanism brittle in exactly the conditions where you need it stable. Keeping them clean, one mechanism for truth and one for accountability, is the correct call.
The Whistleblower's Cut
The validator submitting the slashing evidence doesn't do it out of civic virtue. They receive a reward currently specified as 1/512 of the slashed validator's effective balance, paid immediately from the slashed funds. Not a fortune. But a standing economic incentive to monitor the gossip layer and report slashable conditions promptly.
In practice, most slashing reports are automated. Validator clients like Lighthouse, Prysm, and Teku maintain slashing protection databases and can detect slashable messages without human intervention. The whistleblower reward mostly flows to whichever proposer happens to include the evidence first.
So here's the question worth sitting with: if thousands of self-interested participants are all running automated watchers, all collecting a fee for reporting you, what exactly is the expected lifespan of undetected equivocation on this network?
Vanishingly short. Equivocation is, at its core, a bet that nobody will notice. Gasper's answer is to pay people to prove you wrong.