Platform Overview
A short orientation to Lattix — a distributed management plane, a decentralized data plane, and protected objects that carry their own encryption and policy.
Lattix protects data by combining a distributed management plane, a decentralized data plane, and protected objects that carry their own encryption and policy bindings.
The platform works in three parts:
- Management services define identity, policy, key access, and evidence.
- Mesh nodes distribute policy and enforce it locally.
- Protected objects remain sealed until an authorized unwrap.
Each of the diagrams below tells one part of that story. The five capability layers at the end name the cross-cutting terms the rest of the documentation refers back to.
Platform at a glance
Two planes coordinate; protected objects travel between them. The management plane is where identity providers are wired in, policies are authored, keys are mediated, and evidence is recorded. The data plane is where those policies are replicated and enforced at the node where the request actually happens. The next three diagrams expand the parts that need more detail.
Protected object model
Traditional security architectures protect the container around the data: the server, the network segment, the application. When the data leaves that container, its protection ends. Lattix inverts the model — the envelope travels with the data and every unwrap requires a fresh, policy-authorized key release. Protection applies the same way in your cloud, in a partner's cloud, on a laptop, on a USB drive, or in an email — and revocation remains effective against copies you no longer control.
Authorized unwrap flow
Evidence is generated throughout, not just at the end: policy publication, replica updates, access requests, key releases, and revocations all land on the immutable ledger. Denials produce records too, so the absence of a record really does prove an access did not occur.
How policy reaches every node
Policy is authored once in the management plane and ends up evaluated on every node that needs it, without any node depending on a central evaluator at request time. Revocations and trust updates propagate the same way, so a change at the PAP reaches the enforcing edge within minutes rather than requiring a push-to-every-endpoint step.
The five capability layers
The five capabilities below are not sequential stages — they are cross-cutting layers that appear throughout the diagrams above. They are listed here so the rest of the documentation has stable terms to refer back to.
- /01 Identity and Trust — OIDC, SAML, SPIFFE, mTLS, node identity, attestation. Principals become attribute sets; participants prove themselves with tenant-scoped certificates.
- /02 Policy Authoring and Distribution — PAP for authoring; a common PDP runtime evaluates in both planes; signed policy bundles propagate over the libp2p mesh.
- /03 Protected Object — the TDF envelope itself: ciphertext, wrapped DEK, policy binding, integrity metadata, audit linkage. Persistent, not transient.
- /04 Local Enforcement and Key Access — PEP at the workload, embedded PDP at the node, KAS-mediated unwrap that keeps the KEK inside the trusted key service boundary.
- /05 Evidence and Convergence — immutable ledger, tamper-evident events, revocation sync, policy-version convergence across both planes.
The rest of this documentation
The remaining sections take each capability in turn:
- Core Concepts — the primitives: Trusted Data Format, ABAC, the hierarchical key model, the mesh fabric, the immutable ledger, content addressing, and post-quantum encryption.
- Products — the products that expose these primitives to users: Lattix Passport, Data Rooms, the Zero Trust Fabric, the Mesh Node, the Immutable Ledger, connectors, and the Policy Engine.
- Configuration — the administrative controls a tenant administrator configures.
- Best Practices — recommendations for rolling this out successfully.