Addition and multiplication on encrypted numbers are, by now, textbook homomorphic operations. Comparison is not. Deciding whether one encrypted value is larger than another — without decrypting either — is genuinely awkward, because the schemes that preserve arithmetic structure were not built for the branching logic that “greater than” implies, and the obvious shortcuts leak bits.
US10601579B2, “Privacy preserving comparison,” granted to NXP B.V. on March 24, 2020 and classified under H04L 9/008, addresses exactly this gap. The named inventors include Marc Joye, a well-known figure in homomorphic and lattice cryptography, which is a reasonable signal that the claim is grounded in real scheme design rather than a generic wrapper.
The reason comparison is the hard case is worth making concrete. Homomorphic schemes are good at linear and polynomial operations on the encrypted plaintext. A comparison is a non-linear decision — a step function — and expressing it inside an arithmetic-only world requires either expensive polynomial approximations or an interactive protocol that risks revealing intermediate results. The patent's claim stakes out one disciplined way to get the answer out without getting the inputs out with it.
Why anyone cares: comparison is the atom of almost every useful private computation. Private auctions, encrypted database range queries, privacy-preserving machine-learning decision trees, threshold alerts on encrypted telemetry — all of them reduce to “is this encrypted number above that one?” A scheme that cannot compare cheaply cannot do any of these. So the unglamorous comparison primitive is in fact a gatekeeper for the applications people actually want.
As always on this desk: this is a granted patent, kind code B2, not a pending application, and it describes a method, not a shipped product. NXP is a major secure-element and automotive-security vendor, which is the plausible commercial backdrop, but the patent itself is a claim to a technique.
Read alongside NXP's companion 2020 grant on privacy-preserving evaluation of decision trees, the comparison patent looks less like a one-off and more like a building block in a deliberate private-computation toolkit — the kind of cluster that an IP-strategy reader should note.