Trusted Execution PBFT is a variant of Practical Byzantine Fault Tolerance that leverages Trusted Execution Environments (TEEs), specifically Intel SGX, to implement a Unique Sequential Identifier Generator (USIG) that reduces the protocol to two message phases rather than the standard three. By anchoring sequence number generation inside a tamper-resistant enclave, it eliminates the view-change overhead of classical PBFT while preserving Byzantine fault safety up to f < n/3 faulty replicas.
Trusted Execution PBFT is a variant of Practical Byzantine Fault Tolerance that leverages Trusted Execution Environments (TEEs), specifically Intel SGX, to implement a Unique Sequential Identifier Generator (USIG) that reduces the protocol to two message phases rather than the standard three. By anchoring sequence number generation inside a tamper-resistant enclave, it eliminates the view-change overhead of classical PBFT while preserving Byzantine fault safety up to f < n/3 faulty replicas.
Content
Trusted Execution PBFT addresses the primary performance bottleneck of classical PBFT: its three-phase message complexity and expensive view-change sub-protocol triggered on suspected leader failure. By moving sequence number assignment into a TEE-backed USIG, each replica can verify that a sequence number was assigned exactly once without an extra prepare phase, collapsing the protocol to pre-prepare and commit only.
The Intel SGX enclave provides attestation guarantees: a remote verifier can confirm that the USIG code running on a given processor is exactly the audited implementation and has not been tampered with. This hardware root of trust transforms a distributed trust problem (agreeing on ordering) into a hardware trust problem (trusting the enclave manufacturer), which is an acceptable trade-off in permissioned consortium settings where participants already trust the hardware supply chain.
Trusted Execution PBFT targets permissioned blockchain deployments requiring deterministic finality and high throughput, such as enterprise supply-chain ledgers and financial settlement networks. The reduction from O(n²) to a smaller constant factor in message rounds—combined with the elimination of the costly view-change path—yields measurable latency improvements at moderate committee sizes (10–100 nodes).
Trade-offs include the dependency on Intel SGX availability and its side-channel vulnerability surface (e.g., Spectre, Foreshadow attacks on SGX), as well as the trust assumption on the hardware manufacturer. Ongoing research explores combining TEE-backed USIG with threshold BLS signatures to further reduce per-replica communication cost and to allow dynamic membership without full protocol restart.