Looma: Low-Latency Post-Quantum Authentication for TLS 1.3 in Datacenters

Post-quantum (PQ) authentication in TLS 1.3 can add tens to hundreds of microseconds of handshake processing time. In datacenters, where mutual authentication is mandatory, this authentication cost becomes a dominant contributor to end-to-end request latency, particularly when connections are short-lived and handshake rates are high. This document specifies Looma, an online/offline authentication architecture integrated into the TLS 1.3 handshake. Looma replaces the on-path, per-handshake PQ signature with a fast, one-time signature over the TLS transcript and moves expensive work (including the multi-use PQ signature) to an asynchronous background plane. Looma includes a fallback strategy to preserve correct authentication when the verifier does not have the peer's one-time verification key cached.