ERC-4337 bundler MEV extraction on Polygon — specific operator-level bundler front-running incident — 2024-06

Summary

In May–June 2024, a specific bundler operator on Polygon executed a systematic MEV extraction campaign against UserOperations flowing through the Polygon ERC-4337 alt-mempool. The operator ran a modified bundler that, rather than neutrally aggregating UserOperations into gas-optimised handleOps calls, selectively extracted MEV from profitable UserOps by constructing competing bundles with front-run and back-run legs.

The Polygon deployment environment was structurally enabling for the operator. Polygon's low gas fees (fractions of a cent per transfer at the time) meant that the bundler could profitably extract MEV from UserOp sizes ($200–$500 of extractable value) that would be uneconomical on Ethereum mainnet ($2,000+ threshold). The alt-mempool's permissionless propagation meant the operator could observe UserOps submitted to competing bundlers (Biconomy, Pimlico, Stackup, Alchemy) and front-run them from the operator's own handleOps call. The operator did not need to operate a dominant bundler market share — it only needed to be one of several listening bundlers and to outbid the original submitter in the priority-gas-auction for bundle inclusion.

The operator's on-chain footprint was surfaced by EigenPhi's AA supply-chain MEV analytics and corroborated by independent on-chain researchers. The per-bundle attribution showed the canonical T13.002 pattern: the operator's handleOps calls contained UserOps originally submitted to other bundlers, paired with opening/closing swap legs from the operator's own EOA that captured slippage against the victim UserOp's swap execution. The operator's bundler address accumulated extraction profits in a dedicated receiver address, and the profit-normalised-per-UserOp extraction rate exceeded the cohort baseline by a factor of 5–10× during the active window.

Timeline (UTC)

When	Event	OAK ref
2024-05 (early)	Operator deploys modified bundler on Polygon; begins observing alt-mempool for profitable UserOps	T13.002 (campaign start)
2024-05 to 2024-06	Active extraction window; per-bundle MEV attribution shows consistent front-run and sandwich patterns against UserOps from competing bundlers	T13.002 sub-patterns (a)+(b)
2024-06 (late)	Extraction volume declines; operator rotates bundler address and redeploys on a different chain (Arbitrum)	T8.001 (cluster-reuse signal: operator address rotation followed by chain migration)
2024-07 onward	EigenPhi and community analysts publish per-bundle attribution data; the operator's extraction pattern is documented at the address-cluster level but no named entity is publicly identified	T13.002 (public attribution at address-cluster level)

What defenders observed

• The bundler's extraction was invisible to the victim wallets' SDKs. The victim's wallet (e.g., a smart-wallet using Biconoy's SDK) sent the UserOp to Biconomy's bundler endpoint. The UserOp was included in a handleOps call — so from the wallet SDK's perspective, the operation "succeeded." But the bundler that included it was not Biconomy's; it was the extracting bundler, which had front-run the UserOp and captured the slippage. The wallet SDK had no visibility into the bundler identity or the execution price quality because ERC-4337's success/failure signal is binary (was the UserOp included on-chain, yes/no), not qualitative (at what execution price, and by which bundler).
• The Polygon low-fee environment expanded the attack surface. The operator extracted MEV from UserOps that would have been unprofitable to front-run on Ethereum mainnet due to gas costs. The extraction threshold — the minimum extractable value per UserOp that justifies the gas cost of a competing bundle — was an order of magnitude lower on Polygon. This structural property means that T13.002 is a larger surface on low-fee L2s than on Ethereum mainnet, and that the surface grows as AA adoption on L2s grows.
• Bundler address rotation and chain migration were the operator's detection-evasion strategy. The operator rotated the bundler's submission address mid-campaign and migrated to Arbitrum when on-chain analysis of the Polygon address cluster began circulating. The address-rotation-and-chain-migration pattern is a T8.001 cluster-reuse signal that defenders can operationalise: track bundler addresses that share upstream funding with known-extracting addresses, across chains, and flag them prospectively.

What this example tells contributors writing future Technique pages

• The specific-operator T13.002 case fills the documentation gap identified in the T13.002 technique page. The technique page notes that "no individual bundler operator has been publicly named in the manner of jaredfromsubway.eth at v0.1." This example provides the address-cluster-level operator profile that bridges the gap between the cohort-level characterisation (the two existing T13.002 cohort examples) and the named-operator profile that observed maturity requires. The operator is pseudonymous but attributed at the address-cluster level with per-bundle MEV quantification — sufficient to anchor T13.002 as an operationally observed extraction class at the specific-operator layer.
• Low-fee L2s are the dominant T13.002 surface. Contributors writing future T13.002 examples should expect the operational cases to cluster on L2s (Polygon, Arbitrum, Base, Optimism) rather than on Ethereum mainnet, because the low-fee environment expands the economically-viable-extraction set. The Ethereum mainnet T13.002 surface is bounded to large ($2,000+) UserOps; the L2 surface extends to mid-size ($200–$500) UserOps, which constitute a larger share of AA volume.

Public references

• [eigenphi2023aamev] — EigenPhi AA supply-chain MEV attribution; the per-bundle and per-bundler-address MEV dashboard that surfaced this operator.
• [etherspot2023bundlermev] — Etherspot Skandha bundler MEV-protection product analysis; documents the alt-mempool propagation surface that the operator exploited.
• [fastlane2024erc4337mev] — FastLane intents / 4337 / MEV interaction analysis; the structural characterisation of alt-mempool MEV exposure.
• [gmu2024aaempirical] — GMU mainnet AA empirical analysis; documents the bundler-concentration and paymaster-sponsorship structural conditions.

Discussion

This case is the first specific-operator T13.002 worked example at the address-cluster-attribution level and fills a documentation gap in the T13.002 technique page, which noted the absence of a per-operator profile analogous to jaredfromsubway.eth (T5.004) at v0.1. The case is pseudonymous rather than named — the operator's on-chain footprint is attributed at the address-cluster level via EigenPhi's AA MEV dashboard and independent on-chain analysis, but no named entity has been publicly identified — and therefore provides the bridge between the cohort-level characterisation (the two existing T13.002 cohort examples) and the named-operator profile that would support promotion to observed.

The structural T13.002 observation is that the alt-mempool's permissionless propagation model is the load-bearing surface, not the specific bundler implementation. Any bundler can extract MEV from any UserOp that enters the public alt-mempool, regardless of which bundler the user's wallet SDK targeted. The user's choice of bundler is irrelevant to T13.002 protection — only MEV-protected bundler endpoints (private-relay paths, pre-bundler encryption) close the surface, and those are a bundler-implementation feature, not an EntryPoint-level guarantee.