Skip to main content
Multi-Party Computation (MPC) lets multiple participants compute a function without reconstructing the private inputs on one participant. Protocols differ in trust assumptions, fault behavior, communication cost, and supported operations. Arcium’s current tooling supports Cerberus, a dishonest-majority, detect-and-abort protocol. Cerberus uses secret sharing and authenticated values. Under the assumption that at least one Cluster member is honest, it preserves privacy and aborts when it detects a protocol fault rather than returning a corrupted result. Cerberus uses preprocessing data generated before online execution. This shifts part of the cryptographic work out of the latency-sensitive execution path, but Arx nodes still need sufficient preprocessing material and availability to complete a computation.

​
MPC vs. other encryption techniques

MPC, fully homomorphic encryption (FHE), Trusted Execution Environments (TEEs), and zero-knowledge proofs (ZKPs) solve different problems. Choose based on the application’s trust model and workload rather than treating one technique as universally superior.

​
Fully homomorphic encryption (FHE)

FHE lets a party evaluate supported functions over ciphertexts. It can reduce interaction during evaluation, but computation cost, key management, and supported operations depend on the scheme and workload.

​
Trusted execution environments (TEEs)

TEEs isolate computation in hardware-backed environments. They can provide high performance, but their security depends on hardware, firmware, attestation, and operational controls. MPC distributes trust across participants instead of relying on one enclave.

​
Zero-knowledge proofs (ZKPs)

ZKPs let a prover demonstrate that a statement is true without revealing its private witness. They provide public or designated verification, but do not by themselves let several parties jointly compute over private inputs.

​
Secure multi-party computation (MPC)

MPC is useful when several parties need to compute over private inputs without giving one party all inputs. It requires interaction among participants, so availability and network communication remain part of the system’s liveness assumptions. For a simplified overview, read ELI5: Multi-Party Computation.