Module @ndn/trust-schema

@ndn/trust-schema

This package is part of NDNts, Named Data Networking libraries for the modern web.

This package implements trust schemas.

  • [X] hierarchical trust model verifier
  • [X] hierarchical trust model signer
  • [X] trust schema verifier
    • [ ] override certificate issuer in KeyLocator
  • [X] trust schema signer
    • [ ] choose certificate issuer among existing certificates
    • [ ] automatically create certificates using local issuer key
    • [ ] automatically request certificates from remote certificate authority
import { TrustSchema, TrustSchemaSigner, TrustSchemaVerifier, printESM, versec } from "@ndn/trust-schema";

// other imports for examples
import { Certificate, KeyChain, generateSigningKey } from "@ndn/keychain";
import { Component, Data, ValidityPeriod } from "@ndn/packet";
import assert from "node:assert/strict";

Yingdi Yu proposed trust schema in Schematizing Trust in Named Data Networking. According to his definition:

  • A trust schema comprises a set of linked trust rules and one or more trust anchors.
  • A trust rule is an association of the data name with its signing key name. It can either associate an application data name with its signing key name, or associate a certificate name with its issuer key name.
  • One or more trust anchors, i.e. pre-authenticated keys, are included in the trust schema to serve as bootstrapping points of the trust model.

Pollere LLC released Versatile Security Toolkit (VerSec) as part of Defined-trust Communications Toolkit (DCT). It has a schema description language that describes constraints on:

  • Layout and components of names.
  • Structural and signing relationships between names.

The language specification is fairly complex and contains certain ambiguity. The compiler implementation is found to have several limitations.

python-ndn library authors defined Light VerSec (LVS), a lightweight modification of VerSec that focuses on signing key validation. Its syntax and semantics are similar to VerSec. For ease of processing, LVS introduced some restrictions on identifier names and token ordering. This package can import a subset of LVS models from its textual format via versec.load() function. See @ndn/lvs package for more complete LVS support via its binary format.

The trust schema implementation in this package is inspired by the above documents and projects. The overall structure is:

TrustSchema
+-TrustSchemaPolicy
|   +-patterns = set of
|   |   id => Pattern
|   |         +-ConstPattern
|   |         +-VariablePattern
|   |         +-CertNamePattern
|   |         +-ConcatPattern
|   |         +-AlternatePattern
|   |         \-OverlapPattern
|   |
|   \-rules = set of
|       packet name pattern id => signer name pattern id
|
\-trust anchors = array of Certificate

TrustSchema type represents a trust schema. It contains a TrustSchemaPolicy and an array of trust anchor certificates.

TrustSchemaPolicy type represents the policy portion of a trust schema. It contains a set of patterns, each has a unique id string. It also contains a set of rules, which indicates a packet matching the first pattern should be signed by a key matching the second pattern.

Pattern type represents a pattern in the trust schema policy. It must be one of these sub-types:

  • ConstPattern matches one or more name components specified as a constant in the policy.
  • VariablePattern matches one or more name components (specified as a range), optionally overlapped with an inner pattern and filtered by a JavaScript function. It can save matched components to a variable. When the same variable appears in both packet name pattern and signer name pattern, the matched name component(s) must be the same.
  • CertNamePattern matches either KEY/key-id or KEY/key-id/issuer-id/version suffix in NDN Certificate Format.
  • ConcatPattern concatenates two or more other patterns.
  • AlternatePattern accepts any match among two or more possible patterns.
  • OverlapPattern accepts one or more name components that satisfy two or more overlapped patterns.

This package has partial support of the VerSec syntax, including:

  • component constraints
  • timestamp function: translates to a VariablePattern that matches a Timestamp name component
  • seq function: translates to a VariablePattern that matches a SequenceNum name component
  • sysid, host, uid, pid function: translates to a VariablePattern that assigns to a variable of the same name upper-cased
  • signing constraints and signing chains

Some notes and limitations:

  • This implementation has very limited compile-time schema validation.
  • You can have multiple trust anchors, despite that the VerSec spec allows only one trust anchor.
  • CertNamePattern is created by "KEY"/_/_/_, which should be included at the end of each certificate name.
  • Identifiers starting with _ cannot be used in signing constraints and signing chains.

versec.load() function imports a policy written in VerSec syntax:

const policy = versec.load(`
_site: "a"/"blog"
root: _site/_KEY
#article: _site/"article"/category/year/month <= author

// Notice the variable name distinction between 'adminName' and 'authorName', which is necessary
// to allow them to have different values. Also, the variables cannot be named 'admin' and 'author'
// because that would clash with the pattern names that are implicitly declared as variables.
admin: _site/"admin"/adminName/_KEY <= root
author: _site/_role/authorName/_KEY & { _role: "author" } <= admin

_KEY: "KEY"/_/_/_
`);

versec.print() function prints the policy in VerSec syntax. You may notice that the output differs from the input, because the library has flattened the patterns for faster execution. Occasionally, certain features may not print correctly, especially if the policy was imported from a different syntax.

console.group("VerSec policy");
console.log(versec.print(policy));
console.groupEnd();

printESM() function prints the policy as ECMAScript module. It shows how you can define the same policy in code. However, it cannot automatically convert certain VerSec features, and manual edits would be necessary in such cases. Writing the policy in code can reduce JavaScript bundle size in a web application, because the VerSec compiler is no longer needed at runtime.

console.group("VerSec policy in ESM");
console.log(printESM(policy));
console.groupEnd();

With the policy in place, we can generate a root key and make the trust schema object.

const keyChain = KeyChain.createTemp();
const [rootPvt, rootPub] = await generateSigningKey(keyChain, "/a/blog");
const rootCert = await Certificate.selfSign({ publicKey: rootPub, privateKey: rootPvt });
await keyChain.insertCert(rootCert);
const schema = new TrustSchema(policy, [rootCert]);

TrustSchemaSigner type can automatically select a signer among available certificates in the KeyChain.

const schemaSigner = new TrustSchemaSigner({ keyChain, schema });

const [adminPvt, adminPub] = await generateSigningKey(keyChain, "/a/blog/admin/Lixia");
const adminCert = await Certificate.issue({
publicKey: adminPub,
validity: ValidityPeriod.daysFromNow(30),
issuerId: Component.from("blog"),
issuerPrivateKey: schemaSigner,
});
await keyChain.insertCert(adminCert);
// admin certificate should be signed by root key
assert.equal(adminCert.issuer?.toString(), rootCert.name.toString());

const [authorPvt, authorPub] = await generateSigningKey(keyChain, "/a/blog/author/Yingdi");
const authorCert = await Certificate.issue({
publicKey: authorPub,
validity: ValidityPeriod.daysFromNow(30),
issuerId: Component.from("blog"),
issuerPrivateKey: schemaSigner,
});
await keyChain.insertCert(authorCert);
// author certificate should be signed by admin key
assert.equal(authorCert.issuer?.toString(), adminCert.name.toString());

const articleData = new Data("/a/blog/article/food/2015/1");
await schemaSigner.sign(articleData);
// article should be signed by author key
assert.equal(articleData.sigInfo.keyLocator?.name?.toString(), authorCert.name.toString());

// Data that does not match the policy cannot be signed.
const otherData = new Data("/a/blog/not-article/poison/2015/13");
await assert.rejects(schemaSigner.sign(otherData));

TrustSchemaVerifier type can verify packets according to the trust schema. It can collect intermediate certificates from a local KeyChain and from the network.

const schemaVerifier = new TrustSchemaVerifier({
schema: new TrustSchema(policy, [rootCert]),
offline: true,
keyChain,
});

// The article is trusted.
await schemaVerifier.verify(articleData);

// Although an author could sign the other Data manually, it is not trusted by schema.
await authorPvt.sign(otherData);
await assert.rejects(schemaVerifier.verify(otherData));

References

Renames and re-exports versec