Break a monolithic application into a microservice mesh and you solve some problems and create others. One of the new ones is cryptographic: where there used to be a handful of trust boundaries, now there are hundreds of services, each talking to many others, and every one of those connections should be encrypted and mutually authenticated. That means keys — a great many of them, constantly issued, rotated, and revoked.
US12425380B2, “Secure key management for service mesh deployments,” granted to Intel Corporation on September 23, 2025, addresses that sprawl. Classified under H04L 63/0428 (encrypted network communication) with certificate-based authentication codes, it claims managing keys across the services of a mesh.
The scaling problem is the substance. In a service mesh, mutual TLS between services is the norm, and mutual TLS needs certificates and keys at both ends of every connection. Doing that by hand is impossible; doing it badly — long-lived shared keys, no rotation — recreates the lateral-movement risk the mesh was supposed to contain. Secure, automated, fine-grained key management is what makes a zero-trust service mesh actually zero-trust.
Intel's angle is unsurprising once you consider where it lives in the stack. Intel builds the confidential-computing and hardware-security primitives — enclaves, attestation, hardware key protection — that a mesh's key management can be anchored to. A key-management-for-mesh patent from Intel is a way of connecting its hardware roots of trust to the cloud-native software patterns running on top of them.
Per the desk's discipline: issued grant (B2), not an application; a method/system claim, not a shipped product. Service mesh and confidential-computing tooling is the broad context, and the inventor team (Sood, Addepalli and others) sits in Intel's cloud-security work.
For the systems reader, this is where several threads on this desk converge. Zero-trust segmentation, confidential computing, and key management all meet in the service mesh: it is the concrete architecture where “never trust a connection by default” has to be enforced thousands of times a second, and the keys that enforce it have to be managed at machine speed. Patents like this one are the plumbing under that idea.