Loading...
Loading...
Build cross-organization trust without sharing raw data. DRD's federated trust network connects isolated DRD instances into a unified trust mesh, enabling agent discovery, shared reputation signals, and collaborative governance while preserving each organization's data sovereignty.
Organizations deploying AI agents need to trust agents from partners, vendors, and ecosystem participants. The federated trust network solves this by allowing DRD instances to share trust scores, policy decisions, and agent reputation data without ever exchanging the underlying raw data, event logs, or proprietary policy rules.
Only aggregated trust signals are shared. Raw audit logs, policy rules, and internal data never leave your instance.
Bidirectional trust relationships with configurable sharing levels. Each organization controls what they share.
Data routing respects regional regulations. EU data stays in EU regions, with configurable sovereignty rules.
The federation protocol defines how DRD instances discover each other, establish trust relationships, and exchange trust signals. It operates over mutual TLS connections and uses a publish-subscribe model for real-time trust score updates.
Instances register with the DRD Federation Registry or use DNS-based service discovery (SRV records) to find peers.
Two instances negotiate a federation agreement specifying sharing scope, update frequency, and data retention policies.
Mutual TLS certificates are exchanged and pinned. All subsequent communication is encrypted and authenticated.
Aggregated trust scores are published on a configurable interval (default: 60 seconds). Scores include agent ID, composite score, and confidence level.
Federation health is monitored continuously. Stale connections are flagged, and scores from unresponsive peers are gradually decayed.
The federated knowledge graph connects agents, policies, and trust relationships across organizations into a unified, queryable graph. Each node maintains its own local subgraph and contributes anonymized edges to the federated view.
Contains full agent details, policy rules, event history, and trust score breakdowns. This data never leaves the organization boundary.
Contains agent DIDs, composite trust scores, capability declarations, and relationship edges. All data is aggregated and anonymized before sharing.
Trust signals are aggregated using differential privacy techniques before being shared with federated peers. This ensures that individual events, decisions, or data points cannot be reverse-engineered from the shared scores.
| Technique | Purpose | Privacy Guarantee |
|---|---|---|
| Differential Privacy | Adds calibrated noise to trust scores before sharing | Epsilon-delta privacy budget per query |
| Secure Aggregation | Computes composite scores without revealing individual inputs | No single peer can reconstruct source data |
| Homomorphic Summation | Aggregates encrypted scores from multiple peers | Scores remain encrypted during computation |
| k-Anonymity Binning | Groups agents into trust tiers instead of sharing exact scores | Minimum group size of k=10 per tier |
Discover agents across federated networks using capability-based search, DID resolution, or trust-score filtering. Agent discovery respects each organization's visibility settings -- only agents explicitly marked as discoverable appear in federated queries.
// Discover agents across the federated network
const agents = await drd.federation.discover({
capabilities: ["content-analysis", "image-classification"],
minTrustScore: 80,
regions: ["eu-west-1", "us-east-1"],
limit: 20,
});
// Returns agents from all federated peers
// [
// {
// did: "did:drd:partner-org:019abc...",
// name: "Content Scanner Pro",
// trustScore: 92,
// capabilities: ["content-analysis", "image-classification"],
// organization: "Partner Corp",
// region: "eu-west-1",
// federatedFrom: "partner-corp.drd.io"
// },
// ...
// ]All federation endpoints require mutual TLS (mTLS) authentication. Both the connecting instance and the receiving instance present X.509 certificates signed by the DRD Federation CA or a trusted organizational CA.
DRD automatically provisions, rotates, and revokes federation certificates. Rotation occurs every 90 days with a 30-day overlap for zero-downtime transitions.
Federation peers pin each other's certificate fingerprints during the initial handshake. TOFU (Trust On First Use) with optional out-of-band verification.
Fine-grained controls determine which data stays local and which is shared with federated peers. Sovereignty rules are defined per data category and can be overridden at the workspace, agent, or policy level.
| Data Category | Default | Configurable |
|---|---|---|
| Trust scores (composite) | Shared | Can restrict to local-only |
| Trust score breakdown | Local | Can enable sharing per component |
| Policy rules | Local | Never shared (by design) |
| Audit event logs | Local | Never shared (by design) |
| Agent capabilities | Shared | Can restrict to local-only |
| Agent metadata | Local | Can enable selective sharing |
| Enforcement actions | Local | Can share anonymized summaries |
The federation layer automatically routes data based on jurisdictional requirements. EU data stays within EU regions, and cross-border data flows are governed by configurable rules that map to regulatory frameworks like GDPR, CCPA, and LGPD.
Data pinned to eu-west-1 and eu-central-1. Cross-border transfers require adequacy decision or SCCs.
Data in us-east-1 and us-west-2. Opt-out signals are respected across federation peers.
Data in ap-southeast-1 and ap-northeast-1. China-origin data requires explicit cross-border consent.
Data in sa-east-1. Transfers to non-adequate countries require additional safeguards.
DRD supports multiple deployment topologies for federated networks, from hub-and-spoke to fully meshed configurations. Choose the pattern that matches your organization's compliance requirements and latency targets.
A central DRD instance acts as the federation hub. All peers connect to the hub, which aggregates and redistributes trust scores. Best for organizations with a clear hierarchy.
[Peer A] ──mTLS──> [Hub] <──mTLS── [Peer B]
│
[Peer C]Every DRD instance connects directly to every other instance. Provides the lowest latency and highest resilience but requires O(n^2) connections.
[Peer A] ──mTLS──> [Peer B]
│ \ │
│ \ │
v v v
[Peer C] <──mTLS── [Peer D]Instances within the same jurisdiction form a full mesh, and clusters connect to each other via designated gateway nodes. Balances latency, compliance, and complexity.
┌─ EU Cluster ─┐ ┌─ US Cluster ─┐
│ [A] ── [B] │ │ [D] ── [E] │
│ \ / │ │ \ / │
│ [C] ─────│─mTLS│──> [F] │
└──────────────┘ └──────────────┘// Join an existing federation network
import { DRD } from "@drd-io/sdk";
const drd = new DRD({ apiKey: process.env.DRD_API_KEY });
const federation = await drd.federation.join({
// Federation network to join
network: "enterprise-trust-network",
registryUrl: "https://registry.federation.drd.io",
// Your instance's identity
instanceName: "acme-corp-production",
region: "eu-west-1",
// mTLS certificate (PEM-encoded)
certificate: process.env.DRD_FEDERATION_CERT,
privateKey: process.env.DRD_FEDERATION_KEY,
// Sovereignty controls
sharing: {
trustScores: true, // Share composite trust scores
trustBreakdown: false, // Keep component scores local
agentCapabilities: true, // Share agent capability lists
agentMetadata: false, // Keep internal metadata local
enforcementSummaries: true // Share anonymized enforcement stats
},
// Jurisdiction rules
jurisdiction: {
dataResidency: "eu",
crossBorderPolicy: "sccs-required",
allowedRegions: ["eu-west-1", "eu-central-1"],
},
});
console.log(federation);
// {
// federationId: "fed-019abc-...",
// instanceId: "inst-019def-...",
// status: "active",
// connectedPeers: 12,
// syncInterval: 60,
// joinedAt: "2026-02-14T10:00:00Z"
// }// Query trust scores from federated peers
const trustReport = await drd.federation.queryTrust({
agentDid: "did:drd:partner-org:019abc-def0-7890",
// Include scores from all federated peers
includePeers: true,
// Minimum confidence threshold (0-1)
minConfidence: 0.7,
});
console.log(trustReport);
// {
// agentDid: "did:drd:partner-org:019abc-def0-7890",
// localScore: null, // Not a local agent
// federatedScore: {
// composite: 89,
// confidence: 0.94,
// peerCount: 5,
// lastUpdated: "2026-02-14T09:59:12Z",
// },
// peerScores: [
// { peer: "partner-corp.drd.io", score: 91, confidence: 0.97 },
// { peer: "vendor-inc.drd.io", score: 87, confidence: 0.92 },
// { peer: "platform-co.drd.io", score: 88, confidence: 0.95 },
// { peer: "agency-ltd.drd.io", score: 90, confidence: 0.91 },
// { peer: "startup-xyz.drd.io", score: 89, confidence: 0.93 },
// ],
// aggregationMethod: "weighted-average",
// privacyBudgetUsed: 0.12,
// }
// Use federated trust in a policy decision
const decision = await drd.guard.evaluate({
action: "delegate_task",
context: {
delegateDid: "did:drd:partner-org:019abc-def0-7890",
federatedTrustScore: trustReport.federatedScore.composite,
taskSensitivity: "medium",
},
agentId: "01956abc-def0-7890-abcd-1234567890ab",
});Decentralized Identity
DIDs and Verifiable Credentials that underpin federated agent identity.
Learn morePolicy Engine
Rule-based governance that uses federated trust scores in policy decisions.
Learn moreGenAI Safety
Safety and compliance checks that integrate with federated trust signals.
Learn moreTrust Network
Cross-organization trust scoring and reputation sharing.
Learn more