Skip to content
Tenzro
Security

What is protected, how, and what remains open.

Tenzro roots every key in hardware, pairs every classical signature with a post-quantum one, and bounds every AI agent by terms its owner approved. This page states what each of those means, what stays classical and why, and what is still open. The network is in preview, and governance decides the steps to testnet and mainnet.
Quantum resistance

One guarantee, stated with its limits.

A break of classical cryptography alone (ECDSA over secp256k1 or P-256, Ed25519, X25519) cannot forge a validator vote or certificate, move funds from a Tenzro account, forge a bridge committee attestation, or read a protected secret that Tenzro wraps. The word alone matters: the guarantee holds while the post-quantum primitive of each path holds, and the exceptions are places where the verifier belongs to someone else.
PathStatusWhat it means
Validator votes, certificates, registrationHybridML-DSA-65 with ECDSA P-256 as one composite signature; both legs must verify. ML-DSA-87 with P-384 is the optional validator scheme.
Account transactions and account-signed messagesHybridP-256 and ML-DSA-65 are both required. An account with no ML-DSA-65 key enrolled signs nothing until it enrols one.
Inbound bridge attestationsHybridA committee member counts only with a valid secp256k1 and ML-DSA-65 signature.
Key exchange between nodesHybridPeer connections offer and accept only X25519MLKEM768. A peer that offers X25519 alone cannot complete the handshake.
Consensus channelsHybridX25519 and ML-KEM-768 into HKDF-SHA256, with the handshake signed by composite session keys.
Node RPC authentication (A2A, MCP, MoE)HybridBoth hardware keys sign the connection's TLS exporter. A peer without a verified ML-DSA-65 binding is refused.
Key wraps and TEE key releaseHybridX25519 and ML-KEM-768 into HKDF-SHA256, then AES-256-GCM.
Finality proofsPost-quantumThe default form is a hash-based STARK. An optional Groth16 wrap exists for EVM verifiers and is labeled classical.
Block sync, state sync, gossip, blobs, inference requestsClassical authenticationThese authenticate with the node's Ed25519 key under the hybrid key exchange. A versioned protocol that carries the ML-DSA proof is not built yet.
EVM-standard transactionsClassicalAn Ethereum-style transaction signed with secp256k1 is verified by that standard and carries no ML-DSA-65 leg. Holders who want the guarantee use a native account.
Outbound bridge records, external chains' own signaturesClassicalThe contract or chain on the other side verifies them. A classical ceiling on each mint bounds what a forged source chain can mint.
Hardware attestation rootsClassicalTPM, Secure Enclave and TEE vendors sign with RSA or ECDSA. Tenzro verifies them and cannot change the vendor's algorithm.
Release signingClassicalcosign with a hardware-held key. Published as a classical signature.

The full table, including every key and signature use, is in Post-quantum and AI security.

Harvest now, decrypt later

Where recorded traffic could be read later.

  • Node-to-node connections use the hybrid key exchange only, so recorded peer traffic needs both X25519 and ML-KEM-768 broken to read.
  • Key wraps at rest and TEE key releases use X25519 with ML-KEM-768. The wallet keystore is symmetric (Argon2id and AES-256-GCM).
  • Connections to relays, and RPC and MCP endpoints presented to browsers, use ordinary web certificates. The hybrid exchange is offered first, but X25519 alone is accepted there for a terminator that cannot do the hybrid. That path is the exposure.
  • Some verifiers are outside Tenzro's control: external chains, hardware vendors and standards define their own algorithms, and the residuals below name each one.
Algorithm agility

What can change without a new design.

  • The validator signature schemes a node may register under are a governed bitmask (ML-DSA-65 with P-256, ML-DSA-87 with P-256). Keys already registered are unaffected.
  • The dates from which machine records and session keys must carry the ML-DSA leg are governed values.
  • An identity can rotate its governing key to ML-DSA-65.
  • Key wraps carry an envelope version, so a new wrap format can be introduced beside the old one.
  • The set of known algorithms is fixed in the binary. Adding another signature scheme or a new KEM needs a protocol upgrade, which enacts only after validators holding 67 percent of bonded stake by default have quoted the new binary from their TPMs.
Security for AI agents

An agent is a delegate, not a principal.

The properties below are enforced by chain state and by validators, not by an agent's prompt. Every machine identity is rooted in a TPM, a Secure Enclave or a delegated key holder, and every agent has an owner: a human with a passkey or a machine with a TPM.

A machine without a TPM of its own signs through a machine the same operator owns: the key never leaves that machine's TPM, and the TPM-less machine holds no key, only the right to ask. Accounts survive the loss of a device through linked devices and recovery signers, with a delay that an existing device can cancel.

Supply chain and adversarial testing

Verified before it runs, tested before it ships.

AI tools find vulnerabilities faster than before. Tenzro is maintained on the assumption that weaknesses will be found, and the checks below run on every release or every night.
Pause and guardian levers

What exists, and what does not.

  • Protocol guardians pause the enactment of governance proposals, and governance guardians veto a queued proposal. A pause needs two distinct seated guardians by default and lifts by itself after three days by default. Every guardian power stops things and originates nothing: no guardian action moves funds, sets a parameter or enacts a proposal.
  • Governed status levers (0 enabled, 1 paused, 2 retired) switch individual features, such as inference verification levels, proof systems and recovery kinds, through a normal governance proposal.
  • Validators holding two thirds of stake can pause a bridge route on a finding, and governance can disable or pause one on the expedited track.
  • An agent's owner can pause, quarantine or terminate it, and reclaim its funds.
What this page does not claim

Limits of the evidence.

  • No external security audit has been performed. The threat model scopes that engagement. Nothing on this page is a third-party assurance.
  • No claim that the system is unbreakable. The post-quantum guarantee holds while the post-quantum primitive of each path holds, and breaking both legs of a path breaks the path.
  • No compliance certification is claimed.
  • Hardware attestation proves what the vendor's chain says it proves. A party that forges a vendor chain can claim a machine is genuine hardware; it still cannot sign with that machine's ML-DSA key.
  • Tooling that scans for vulnerabilities, including AI tools, is a way to find defects. A scan with no findings is not evidence of absence, and none is cited here as assurance.
Open items

Known residuals, stated plainly.

These are listed in the project's own residuals notes. Each is open until the repository says it is closed.
  1. 01
    Machine records still accept a missing ML-DSA leg.
    The governed switch records.pq_leg_required_from_ms defaults to never, and the launch genesis in the current source does not set it. Usage records, meter bond consent, reserve attestations, sensor readings, batch roots, session keys and TEE identity evidence verify under a policy that checks the ML-DSA-65 leg only when the key has one, until it is set.
  2. 02
    Other verifiers check the ML-DSA-65 leg only when the key has one.
    These include sealed-model manifests, identity envelopes, session certificates, the admin audit log and the read-only signature check.
  3. 03
    Authentication on block sync, state sync, gossip, blobs and inference requests is classical.
    Their encryption uses the hybrid key exchange; their peer authentication uses an Ed25519 key.
  4. 04
    Duty handover proofs and TEE finality attestation bind a Groth16 seal.
    A STARK-verified duty needs a consensus rule and a change to the finality program. Binding a STARK to a TEE report needs the report-data layout changed.
  5. 05
    Wrong-handover evidence for bridge committees checks only the secp256k1 signature.
    A party that forges secp256k1 signatures could frame a committee member until the handover format carries both legs.
  6. 06
    The node's own DID document proof is a P-256 proof.
    The ML-DSA-65 method is listed in the document the proof covers, but the document carries no ML-DSA-65 signature of its own.
  7. 07
    The release signing key's fingerprint is not published independently of the release page.
    Until it is, compare the SHA-256 digest in the release notes as a second reference. Signatures are not recorded in a transparency log. Windows archives are not Authenticode-signed unless the build host had a certificate.
  8. 08
    The participation roster at epoch close must be redesigned before mainnet.
    A node that lacks the closing epoch's validator set skips a write the others make, and its state root differs. Low risk on a freshly launched network.
  9. 09
    The record age rule has no activation gate.
    A later upgrade of a running network must take effect from a chain height.
  10. 10
    A restarted validator can stay outside the mesh for minutes under packet loss.
    A repeat of the sealed endpoint set shortens it. A four-validator run that stopped finalizing for 246 seconds is not explained, because its logs were not kept.
  11. 11
    Bridge routes without a light client rest on their committee.
    Their safety is bounded by the caps, the breaker and the supply invariant. A bridge signer backend that reads a key from the environment still exists for development.
Responsible disclosure

Report a vulnerability.

If you believe you have found a security issue in Tenzro Network, email eng@tenzro.com or use GitHub's private vulnerability reporting on the affected repository. Please do not open a public issue, and do not test against other people's accounts or nodes. We acknowledge each report, keep you informed while we work on the fix, and agree the disclosure with you. Security fixes go into the latest release.
ExploreJoin the network