Skip to content
Crypto.world

Research

Enhancing the Security of Rollup Sequencers using Decentrally Attested TEEs

Rollups scale blockchains by moving transaction processing off-chain, but they depend on a Sequencer — a privileged, centralized node that orders and batches transactions — which can censor, reorder…

Enhancing the Security of Rollup Sequencers using Decentrally Attested TEEs
arXiv
2511.22317
Published
2025-11-27
Authors
Giovanni Maria Cristiano, Salvatore D'Antonio, Jonah Giglio, Giovanni Mazzeo, Luigi Romano

AI summary

Overview

Research area: Blockchain Layer-2 scaling and systems security — specifically securing the Rollup Sequencer using Trusted Execution Environments (TEEs) and decentralized remote attestation.

Technical level: Advanced. The paper assumes familiarity with Rollup architecture (Optimism/OP Stack), Intel SGX/TDX, DCAP remote attestation, and smart contract design, although the core idea can be grasped at a conceptual level.

Scope (one sentence): The paper designs, implements, and benchmarks a Sequencer that runs inside an Intel SGX enclave and whose attestation quote is verified by on-chain smart contracts rather than a centralized attestation service, evaluated on a locally built Optimism-based Rollup testbed.

What This Paper Is About

Rollups scale blockchains by moving transaction processing off-chain, but they depend on a Sequencer — a privileged, centralized node that orders and batches transactions — which can censor, reorder, delay, or tamper with user transactions. Prior work has shielded the Sequencer (or the whole Rollup) inside a TEE, but TEE attestation itself typically relies on centralized infrastructure (Intel's Provisioning Certification Service and a Provisioning Certificate Caching Service), which reintroduces a single point of trust. This paper's goal is to protect only the Sequencer inside a TEE while moving quote validation and policy checks onto smart contracts, so anyone can verify the Sequencer's integrity without a trusted intermediary.

Key Contributions

  1. Targeted TEE protection. The authors isolate only the Sequencer (the op-node and op-geth components of the OP Stack) inside a TEE, rather than the entire Rollup infrastructure, avoiding the complexity and overhead of full-Rollup enclaving.
  2. Decentralized attestation mechanism. They design and implement on-chain verification of TEE quotes using smart contracts, removing dependence on a centralized PCCS while acknowledging that Intel's PCS remains the hardware root of trust.
  3. Practical implementation. They build the system using Intel SGX (with TDX noted as sharing the same attestation process), the Optimism protocol, the Gramine library OS, and libraries adapted from Automata (Automata-dcap-attestation and Automata-dcap-qpl), written in Solidity 0.8.0.
  4. Comprehensive evaluation. They run an experimental evaluation on a locally developed Optimism-based Rollup testbed, measuring throughput, latency, CPU and memory overhead, and the cost/feasibility of the on-chain attestation process.

Main Findings

  • Throughput cost: The TEE-protected Sequencer achieves 7–8 TPS versus 20–25 TPS in the native (unprotected) configuration.
  • Latency cost: Confirmation latency increases from 2.5–3.5 seconds in the native configuration to 21–25 seconds with the TEE-protected Sequencer.
  • Resource overhead: CPU usage increases by approximately 60–80% and memory usage by 30–50% compared with the native setup.
  • Security gain: Despite these overheads, the Sequencer gains hardware-enforced integrity and confidentiality, becoming resistant to transaction manipulation, censorship, and other host-level attacks possible in the unprotected setup.
  • Attestation feasibility: The decentralized attestation mechanism was benchmarked and confirmed to be feasible, with manageable on-chain verification costs (specific gas figures are not reported in the provided text).
  • Renewal design trade-off: A 4-hour quote validity window was selected as balancing security and gas expenditure; the Sequencer proactively renews when remaining validity drops below 20%.
  • Anti-abuse parameters: External renewal requests are restricted to whitelisted or governance-approved addresses, rate-limited (for example, a minimum interval of 30 minutes), require a reason code, and require a small bond (for example, 0.1 ETH) that is slashed if the current attestation still has 50% remaining validity.
  • Design choice of SGX: Intel SGX was chosen over TDX because it offers a smaller Trusted Computing Base and typically higher performance overhead, making the evaluation a deliberate worst case.
  • Porting effort: Running the Sequencer under Gramine required replacing unsupported system calls (for example, replacing mremap() with mmap() and sendmmsg() with a loop over sendmsg()) and removing non-essential ones (for example, inotify_init1()), as summarized for op-geth in the paper's Table III.

Methodology in Plain English

The researchers started from the standard Optimism Rollup design and changed two things.

First, they moved the Sequencer — the pair of programs that receive, order, and execute transactions — inside an Intel SGX enclave using Gramine, a library OS that lets mostly unmodified applications run in a TEE. Because Gramine does not support every system call the Go-based Optimism components use, they rewrote or removed the incompatible parts until the Sequencer ran correctly inside the enclave.

Second, they built an attestation pipeline on-chain. While running, the Sequencer produces a cryptographic quote that bundles standard SGX measurement fields (such as MRENCLAVE and MRSIGNER) with Rollup-specific values: block hash, block height, state root, L1 origin reference, timestamp, a Sequencer nonce, and the Prover's public key. A Collateral Retriever contract fetches the supporting certificates from Intel's Provisioning Certificate Service and stores them on-chain. The quote is then submitted to an Entry Point Verifier contract, which routes it to a Quote Verifier contract that checks the signature, the structure, the collateral (PCK certificates, revocation lists, TCB status), the freshness window, and whether the embedded metadata matches an on-chain policy. If everything passes, the attestation registry marks the Sequencer as authorized for a limited period. The op-batcher and op-proposer consult this registry and refuse to publish batches or state roots when the Sequencer has no valid, recent attestation.

To test the result, they built a local Rollup testbed based on Optimism and compared the protected and native configurations on throughput, confirmation latency, CPU usage, and memory usage, and separately benchmarked the on-chain attestation path.

Why This Matters

Impact on research: The paper carves out a middle ground between "trust the operator" and "enclave the entire Rollup," showing that protecting a single critical component can deliver meaningful integrity guarantees. It also offers a concrete pattern for replacing centralized attestation caching services with smart contract logic, which is directly relevant to the broader decentralized-TEE literature (DHTee, OPERA, JANUS, Automata).

Real-world applications:

  • Decentralized exchanges and DeFi on Rollups: reducing the risk of front-running, sandwich attacks, and selective censorship by an exchange's underlying Sequencer.
  • Fair ordering services: providing publicly verifiable evidence that transactions were ordered by attested code rather than by an operator's private interest.
  • Rollup-as-a-service and L2 operators: offering a deployable add-on that strengthens trust in a Sequencer without redesigning the Rollup stack.
  • Auditing and monitoring: letting validators, users, or monitoring services trigger attestation renewals and detect lapses in Sequencer integrity on-chain.

Industry relevance: The design deliberately preserves the existing centralized Sequencer model used in production Optimism deployments, so it is framed as an incremental hardening step rather than a rewrite — important for teams that cannot yet decentralize sequencing. The quantified costs (7–8 TPS versus 20–25 TPS, 2.5–3.5 s versus 21–25 s latency, plus CPU and memory overhead) give operators a concrete basis for deciding whether the guarantee is worth the price.

Future Directions

  • Closing the performance gap. The measured drop to 7–8 TPS and 21–25 second latency raises the question of how much of the overhead comes from Gramine, from SGX itself, or from the attestation flow, and whether TDX or a lighter library OS would narrow it.
  • Removing the remaining trust anchor. The authors concede that Intel's PCS cannot be decentralized because it is the hardware root of trust; whether multi-vendor TEEs or alternative roots could reduce this dependency is left open.
  • Handling multiple Sequencers. Coordination among multiple Sequencers and multi-Sequencer MEV strategies are explicitly declared out of scope, leaving decentralized sequencing combined with attestation as an open design problem.
  • Hardening against lower-level adversaries. Hardware attacks, SGX side-channels (including speculative execution exploits such as Spectre and Foreshadow), verification-contract bugs, and Layer-1 consensus compromise are listed as out of scope or residual risk, marking natural extensions for follow-up work.

Target Audience

This paper is most valuable to Layer-2 protocol engineers and architects working on Optimism-style Rollups, to confidential-computing researchers applying Intel SGX/TDX and remote attestation to blockchain systems, and to security engineers evaluating the practical cost of enclaving performance-critical blockchain components. Graduate students and researchers studying trust minimization in Rollups will also find the threat model, the on-chain attestation design, and the measured overheads a useful reference point.

Authors’ abstract

The growing scalability demand of public Blockchains led to the rise of Layer-2 solutions, such as Rollups. Rollups improve transaction throughput by processing operations off-chain and posting the results on-chain. A critical component in Rollups is the Sequencer, responsible for receiving, ordering and batching transactions before they are submitted to the Layer-1 blockchain. While essential, the centralized nature of the Sequencer makes it vulnerable to attacks, such as…

The full abstract on arXiv
Read the original paper