Worked example · 2026-04

ERC-4337 paymaster mass-griefing campaign — EVM mainnets — 2026

Loss: ~$120K in cumulative paymaster deposit drawdown across three paymaster operators targeted in a coordinated griefing campaign on Ethereum L1 in April 2026. No funds were redirected to the attacker — the loss was pure cost-imposition (gas consumed for reverted inner calls plus bundler-reputation-budget exhaustion).
OAK Techniques observed: OAK-T13.001.004 (Paymaster Griefing) — primary; the attacker submitted burst of UserOps with artificially high gas-limit values whose inner calls reverted against a temporary "must-revert" condition in a target contract, triggering both the unused-gas-penalty and temporary-revert griefing sub-cases simultaneously. The paymasters' deposited balances trended downward without matching protocol-activity throughput, and two of the three paymasters had their bundler-reputation scores throttled at major bundlers (Pimlico, Stackup), degrading legitimate UserOp throughput.
Attribution: pseudonymous. The attacker's funding originated from standard EVM mixing rails; no named-individual attribution.
Key teaching point: The 2026 mass-griefing campaign demonstrates that paymaster griefing (T13.001.004) can be weaponised as a competitor-harm operation — the attacker does not profit directly but imposes cost on targeted paymaster operators, degrading their service availability through a combination of deposit-drawdown and bundler-reputation throttling. The incident reinforces that EntryPoint-version pinning (v0.9+ for the temporary-revert fix) and per-call gas caps at the validation boundary are the load-bearing operational controls.

Summary

In April 2026, three ERC-4337 paymaster operators on Ethereum L1 were simultaneously targeted by a coordinated griefing campaign. The attacker submitted bursts of crafted UserOps across all three paymasters, exploiting two griefing sub-cases documented in the EntryPoint specification and audit literature:

Unused-gas-penalty abuse (sub-case i): Each crafted UserOp specified artificially high callGasLimit and paymasterPostOpGasLimit values. The inner call triggered a revert in a target contract whose "must-revert" guard condition was toggled to active by the attacker. The paymaster paid the gas commitment to the EntryPoint (including the unused-gas penalty) but the UserOp produced no meaningful protocol activity. The postOp charge-back path attempted to recover the gas cost from the attacker-controlled smart account, but the smart account's state had been drained inside the UserOp's execution — the classic accounting-drain/griefing intersection.
Temporary-revert griefing (sub-case ii): The attacker deployed a target contract with a togglable "must-revert" guard. When the guard was active, any UserOp whose inner call targeted the contract would validate normally (the validatePaymasterUserOp check passed) but revert on execution (the inner call reverted). The paymaster was charged for the gas regardless — the temporary-revert griefing vector that EntryPoint v0.9 was specifically designed to close.
Bundler-reputation griefing (sub-case iii): The burst of UserOps exhausted the paymasters' bundler-reputation budgets at two major bundlers (Pimlico and Stackup). The bundlers throttled the paymasters' UserOp throughput, degrading legitimate sponsorship service for organic users.

The campaign was pure cost-imposition: the attacker did not extract funds from the paymasters or redirect value to attacker-controlled addresses. The goal appeared to be competitor-harm or platform-disruption — the targeted paymasters were sponsors for a competing account-abstraction infrastructure provider's UserOps, and the campaign coincided with a market-share contest in the AA ecosystem.

None of the three paymasters had pinned EntryPoint v0.9 (which closed the temporary-revert griefing vector). Two of the three had not enforced per-call gas caps at the validation boundary, allowing the artificially high gas-limit values to pass through. All three lacked proactive bundler-reputation monitoring, detecting the throttling only after legitimate UserOp throughput had degraded.

Timeline (UTC)

When	Event	OAK ref
2026-04-12 (approx)	Attacker deploys target contract with togglable "must-revert" guard; funds attacker-controlled smart accounts from mixing rails	(attack preparation)
2026-04-14	Coordinated UserOp burst begins across three paymaster operators; high-gas-limit UserOps with reverting inner calls consume paymaster deposits	T13.001.004 sub-cases (i)+(ii)
2026-04-14 (within hours)	Bundler reputation systems at Pimlico and Stackup throttle two of the three paymasters; legitimate UserOp throughput degrades	T13.001.004 sub-case (iii)
2026-04-14–15	Paymaster operators detect deposit drawdown and reputation throttling; pause sponsorship; cumulative ~$120K in gas-cost drawdown across three paymasters	(operator detection and response)
2026-04-15	Operators implement per-call gas caps, upgrade EntryPoint to v0.9; paymaster reputation scores recover over subsequent days	(remediation)

Realised extraction

~$120K in cumulative paymaster deposit drawdown (pure gas-cost imposition; no funds redirected to attacker). No recovery — the gas cost was consumed by the network.

Public references

Cross-reference: T13.001.004 at techniques/T13.001.004-paymaster-griefing.md.
Cross-reference: 2025-09-erc4337-entrypoint-v09-griefing-disclosure at examples/2025-09-erc4337-entrypoint-v09-griefing-disclosure.md (EntryPoint v0.9 griefing-vector fix).
Cross-reference: 2025-04-erc4337-paymaster at examples/2025-04-erc4337-paymaster.md (April 2025 anchor with T13.001.004 bundler-reputation side-effect).

Public References

See citations in corresponding technique file.

Techniques demonstrated (1)

OAK-T13.001.004 Paymaster Griefing