GitHub Devnet Live

The Uruz DAG.
In detail.

A deep dive into the protocol mechanisms behind URUZ — from DAG structure and work-weighted finality to post-quantum checkpoints, adaptive parameters, and Structural Graph Identity.

Layer 1 · DAG Architecture

A graph, not a chain.

Traditional blockchains serialize transactions into a linear chain. URUZ uses a Directed Acyclic Graph where each vertex references multiple parent tips simultaneously, allowing concurrent emission by all participants without coordination.

The DAG naturally handles forks: competing vertices are not discarded but accumulated. Finality emerges from work accumulation in descendants, not from a committee choosing a winner.

The protocol is implemented in Rust with a modular architecture focused on deterministic behavior, throughput, and operational resilience.

Protocol model (conceptual)
// Each DAG vertex references prior vertices and contributes work. // Finality state progresses only when bounded policy conditions are met. // Safety conditions are deterministic and independently verifiable.
Adaptive parameter profile — current devnet class
Safety floor (technical)
Prevents under-constrained operation
active
Safety floor (economic)
Protects against low-cost influence spikes
active
Depth minimum
Maintains irreversibility confidence
policy-bound
Delay smoothing
Stabilizes adaptive responses over noisy conditions
dynamic
Sampling window
Robust telemetry buffer for adaptive estimation
dynamic
Anomaly threshold
Guards against abrupt adversarial parameter shifts
active
Recompute cadence
Bounded update frequency for network stability
bounded
Warm-up profile
Controlled startup behavior before full adaptation
enabled
Adaptive parameters

Parameters that adapt. Floors that don't.

Finality parameters adapt to network conditions in real time while remaining bounded by strict safety constraints.

Adaptive parameters are estimated from robust network telemetry with outlier resistance and bounded-rate updates.

Startup warm-up logic keeps thresholds stable during node restart and early sampling.

Adaptive control loop (conceptual)
// Measure network conditions // Apply bounded smoothing and safety floors // Reject anomalous updates
Post-Quantum · FIPS 204

Quantum-resistant from day one.

Every checkpoint in URUZ is certified using ML-DSA (Module-Lattice Digital Signature Algorithm), standardised as FIPS 204 — the same algorithm NIST selected as its primary post-quantum signature standard in 2024.

Checkpoint verification is rooted in immutable trust anchors. Snapshot import requires strict integrity checks to prevent poisoned state acceptance.

Bootstrap acceptance requires independent peer corroboration plus anti-abuse rate limiting.

Checkpoint anchoring

Three-layer bootstrap integrity.

Deployed in current, the bootstrap protection operates in three layers:

Layer 1 — immutable trust anchor verification.

Layer 2 — multi-source consistency checks before accepting state.

Layer 3 — controlled recovery cadence to prevent abuse.

Bootstrap verification (conceptual)
// Validate anchor lineage // Require corroboration from independent peers // Enforce controlled recovery cadence

Sybil resistance from behavior.

Reputation is derived from verifiable patterns in the DAG itself — not from external capital, hardware, or identity.

Reputation function
Reputation is derived from durable graph behavior, diversity, and consistency signals under bounded influence rules.
Work accumulation in the DAG
Diversity-aware referencing behavior
Checkpoint-linked consistency over time
Operational reputation governs relay quality
Consensus reputation is activated in phases

Consensus reputation is intentionally slow-moving, so short-term bursts cannot capture meaningful influence.

Diversity filters reduce coordinated behavior inflation and keep influence tied to independent participation quality.

Finality influence transitions are smoothed across windows to prevent one-cycle manipulation.

Phase 0
Flat counting
SGI in observe-only shadow mode with data collection for calibration.
Phase 1
Hybrid mode
Reputation influence starts bounded and expands only after objective maturity and safety gates are met.
Phase 2
Full SGI
Consensus reputation drives full finality influence only after formal go/no-go criteria are met.

Defense in depth.

The security specification defines a structured threat model with phased defenses, telemetry, and response policies.

Reputation threats
Reputation attacks
Coordinated influence inflation attempts are mitigated with staged reputation and diversity safeguards.
Mitigated by phased controls
Protocol threat
Sensor poisoning
Telemetry manipulation attempts are mitigated through robust signal filtering and anomaly defenses.
Mitigated by telemetry safeguards
Protocol threat
Eclipse / partial-view
Partial-view and partition conditions are mitigated with network diversity and checkpoint consistency defenses.
Mitigated by network and consensus safeguards
Protocol threat
Concurrent checkpoints
Competing checkpoint behavior is mitigated with quorum validation and participation diversity safeguards.
Mitigated by quorum and diversity safeguards
Protocol threat
Poisoned bootstrap
Poisoned bootstrap attempts are mitigated with anchored verification and multi-source consistency checks.
Mitigated by anchored bootstrap validation
Protocol threat
Validity record poisoning
False-attestation attempts are mitigated through attestation integrity checks and challenge procedures.
Mitigated by attestation integrity safeguards

Ready to build on URUZ?

Join Testnet Waitlist View Roadmap