Stats · the numbers behind the proof

Verifiable. Not estimated.

Everyone publishes a dashboard of "best estimates." Ours is a tamper-evident, Merkle-rooted ledger — every number below is one you can recompute yourself.

$2.03B
Re-proven exploit value
public loss of the incidents we re-executed
21 / 32
Historical exploits re-proven
replayed on a fork at the original block
141
Vulnerability classes proven
each one runs on a real fork
0
Proven false-positives
on our audited baseline

The per-receipt transparency log is empty: the engine's signed receipts are not being appended to it yet, so the command below has nothing to verify and the published root is the genesis value. The numbers above do not depend on it — each is reproducible from the artifacts in the repository.

Findings over time

Each point is a signed receipt appended to the public proof log. The count climbs as the finding factory runs.

Cumulative findings

By chain

Audit this number

The claim above is only as good as its verifiability. Here's how to recompute it — no trust required.

The proof log is append-only and hash-chained (Sigstore/Rekor model). Fetch it, verify the chain, and recompute the Merkle root — it must match what we publish:

# dependency-free — no trust in us required:
curl -s https://omniguardlabs.com/proofs/verify_root.py -o verify_root.py
curl -s https://omniguardlabs.com/proofs/proof_log.jsonl | python3 verify_root.py -

Published Merkle root: —

Mutating, reordering, or removing any past entry changes the root. That's the point — no estimate, a ledger. See the Proof Wall →

What these numbers do not claim

Honesty is the product. So the fine print is loud.

Findings = execution-proven, signed, per-instance — never heuristic flags Classes = taxonomy; findings = instances — never conflated Re-proven value = the public loss of incidents we re-executed ourselves, not "value we secured" A queued incident's loss is never counted as work performed Re-proven from DeFiHackLabs + Immunefi/PeckShield incidents Chains proven: —