Secure multiparty computation gets described as if the computation is the whole story: several parties jointly compute a function of their private inputs while keeping those inputs secret. The cryptography literature lavishes attention on the computation — the protocols, the rounds, the security proofs. The input stage tends to be waved away as setup. It is not.
US11625490B2, “Method and apparatus for obtaining input of secure multiparty computation protocol,” granted in April 2023 and classified under H04L 9/008 with multiparty-computation codes (H04L 2209/46), takes the input stage seriously as a problem worth its own claim.
The reason it deserves attention is that an MPC is only as trustworthy as the inputs fed into it. If a party can supply a malformed or maliciously crafted input, it can skew or corrupt the jointly computed result even while the computation protocol itself runs flawlessly. Validating and committing inputs — binding each party to a well-formed value before the computation proceeds — is therefore part of the security boundary, not a preliminary nicety.
There is also a confidentiality dimension. Each party's input must enter the protocol without being revealed to the others, which is itself a small cryptographic problem: encode and commit a secret value, prove it is well-formed, and make it available to the computation without exposing it. The patent's framing — “obtaining the input” — is precisely this under-discussed handoff.
Per the desk's discipline: issued grant (B2), not an application; a method/apparatus claim, not a product. The assignee sits in the broader ecosystem of operating companies (the Ant/Alibaba-linked Advanced New Technologies) that have filed heavily in privacy-preserving computation, which is the commercial context.
For the reader trying to evaluate MPC claims, the lesson is to look past the headline computation. The robustness of an MPC system lives in the boundary steps — input commitment, output reconstruction, abort handling — and patents that stake out those steps are quietly load-bearing, because every real deployment has to solve them.