Fully homomorphic encryption earns the word “fully” through one operation: bootstrapping. Every homomorphic computation adds noise to the ciphertext, and after enough operations the noise would swamp the data. Bootstrapping refreshes the ciphertext — reducing the noise so computation can continue indefinitely. It is also, by a wide margin, the most expensive thing FHE does.
US12143466B2, “Interactive bootstrapping for approximate homomorphic encryption,” granted to Duality Technologies, Inc. on November 12, 2024, attacks that bottleneck. Classified under H04L 9/008, it claims an interactive approach to bootstrapping — using a round of interaction to make the refresh cheaper than the fully non-interactive version.
The inventor list is a who's-who: Micciancio, Polyakov, and Vaikuntanathan are foundational figures in lattice cryptography and FHE. That lineage matters for credibility — this is not a generic optimization claim but work from people who built the schemes being optimized, and it tags “approximate” homomorphic encryption, the CKKS-style arithmetic that dominates encrypted machine-learning workloads.
The trade the claim makes is the interesting part. Non-interactive bootstrapping is self-contained but brutally expensive; allowing a controlled round of interaction can dramatically cut the cost, at the price of needing the parties to exchange messages. For many deployments — a client and a server in an ongoing session — a little interaction is entirely acceptable if it makes the refresh affordable. The patent stakes out that trade.
Per the desk's discipline: issued grant (B2), not an application; a method claim, not a shipped product — though Duality's OpenFHE library and commercial FHE platform are the obvious context. Duality appears repeatedly in the 2024 FHE record, including a companion grant on large-precision homomorphic comparison via bootstrapping.
For the strategy reader, the pattern is clear: the FHE frontier in 2024 is bootstrapping efficiency, and a specialist firm staffed by the field's founders is fencing the techniques. Whoever owns affordable bootstrapping owns affordable FHE, because every long computation has to refresh, and refreshing is where the cost lives.