Methodology

We gather the whitepaper, protocol specs, source code, testnet data, and public incident reports. The goal is to understand what the protocol claims and what it actually depends on.

We build a Protocol Digital Twin: a deterministic, interactive model of the network, consensus, cryptography, and execution layers. The twin is the shared reference for every later test.

Signatures, hashing, randomness, key derivation, and zero-knowledge circuits are reviewed against their stated assumptions. We look for weak curves, bias, replay patterns, and implementation gaps.

Safety, liveness, and finality are tested under network partitions, validator crashes, byzantine behaviour, and clock skew. We verify quorum math and timeout parameters.

We run adversarial scenarios from the attack library: Sybil, Eclipse, Partition, Leader Failure, Predictable Randomness, Resource Exhaustion, and custom protocol-specific attacks.

Throughput, latency, finality, resource use, and storage growth are measured under normal and adversarial conditions. Performance numbers are only meaningful when paired with correctness.

Findings are triaged, risk-scored, and translated into actionable remediation. Residual risk is documented. The twin remains available for regression testing after fixes are deployed.