Skip to content

Hand an auditor a signed Wazuh report

Machine verified A machine ran this on our install and checked the result against the CLI. A test report is linked. Proven against quoxcli 0.5.0 on 2026-11-10.

Proof for the auditor

A posture report signed with a published key. Change one number and it stops verifying.

What you get: a stored report covering the Wazuh manager, agent health, the exploit-aware patch ranking, recent alerts and compliance scores, signed with your install’s WARD key. The auditor checks it offline, with the public key and nothing else.

quox · talking to Quox

Produce a signed Wazuh security posture report I can hand to our auditor. Give me its id, and tell me how the auditor can check it hasn't been altered without trusting us. Don't change anything on the servers.

▤ AOCL │ L1 identity complete · allowed · 1ms

▤ AOCL │ L3 policy gate complete · proceed · 11ms

▤ AOCL │ handoff narration kept out of the answer · I'll hand this to ARGUS, our Wazuh specialist, as a read-only request.

◐ DELEGATE │ delegate_to_agent agent_id=argus ✔ DELEGATE │ delegate_to_agent agent_id=argus │ 3 tools used, 3 tool outcomes · 40.8s

◐ ARGUS › WAZUH │ posture report ✔ ARGUS › WAZUH │ posture report │ 2.2s

◐ ARGUS › WAZUH │ agents ✔ ARGUS › WAZUH │ agents │ 92ms

◐ ARGUS › WAZUH │ sca ✔ ARGUS › WAZUH │ sca │ 75ms

⛓ AEE │ envelope opened · ops.tool · delegate_to_agent ok

⛓ AEE │ envelope opened · stream_success

I've produced the signed report and it is stored. None of the reads failed and nothing
changed on any server. The tool did not return the signature bytes in its output. They
are only in the exported bundle, so I can't quote the signature here.


Report identity and signature

• Report id:  WPR_MW9YNNFH_98B591BF7D , generated 2026-11-10T00:37:30.845Z, covering the
  whole fleet (source: wazuh_posture_report)
• Report hash: SHA-256 of the canonical JSON:  
  a8b41aff23317f7c5e88561bce87db0b20d7f6dd5e01b419d78e1997bc6c889d
• Signing method: ed25519, using this install's persistent WARD key. The report is
  marked as anchored to that key ( trust_anchored: true ).
• WARD key id:  f695fbcf6be7179b 
• WARD public key (base64):  9pX7z2vnF5ufAMjQtke0VogdIL/2PcdpQ6NMljWcVaM= 
• Where the signature is: in the exported bundle, at  signature.signature  (base64). It
  signs  signature.message , a canonical JSON statement that ties together the report
  id, org, scope, time and  report_sha256 .
• Hashing rule: the hash covers the UTF-8 bytes of canonical JSON (VOLT SPEC 3.2).
  Object keys are sorted at every level using the JavaScript default sort. Keys whose
  value is undefined are dropped. Arrays keep their order. There is no whitespace. Every
  scalar is written exactly as JSON.stringify writes it.
• Hash chain: the tool returned no chain reference. The only link is the report hash
  bound into the signed message.


How an auditor verifies it without trusting Quox

1. You export the bundle:  quox wazuh report show WPR_MW9YNNFH_98B591BF7D --out 
   posture.json
2. You give the auditor  posture.json . Send the WARD public key separately through a
   different channel. They must not take it from the file.
3. The auditor saves the key as  ward-pub.pem  and pins it.
4. The auditor runs  quox wazuh report verify posture.json --public-key ward-pub.pem .
   If the report was changed or signed with a different key, the check fails.
5. For a check that uses no Quox tooling at all, the auditor can:
   ◦ rebuild the canonical JSON of the report using the rules above;
   ◦ compute its SHA-256 and compare it with  report_sha256 ;
   ◦ confirm the hash matches the value in  signature.message ;
   ◦ verify  signature.signature  over  signature.message  with any standard ed25519
     library and the pinned key.

The online check is  quox wazuh report verify WPR_MW9YNNFH_98B591BF7D , but it relies on
Quox. The steps above do not.


What the signed report says

• Manager: 2 agents active, 0 disconnected, never connected or pending (source:
  wazuh_posture_report).
• Coverage: 0 disconnected agents and 0 agents missing a vulnerability inventory. There
  is no server inventory registered, so the report cannot show which hosts lack an agent
  (source: wazuh_posture_report).
• CVEs: 2002 of 8062 were ranked. 7 are on the CISA KEV list and 10 have an EPSS score
  above 0.1. Both feeds answered. In the order the report ranks them, the ones to patch
  first are:
  1. CVE-2026-24061: on KEV, EPSS 0.99, CVSS 9.8, 1 host.
  2. CVE-2026-31431: on KEV, EPSS 0.98, CVSS 7.8, 2 hosts.
  3. CVE-2026-111497: on KEV, EPSS 0.83, 1 host.
  4. CVE-2026-46300: on KEV, EPSS 0.16, 2 hosts.
  (source: wazuh_posture_report)
• Alerts in the last 24 hours: 8 in total.
  ◦ Rule 100512, level 12, 2 hits. Its description reads "DEMO (Quox proof): sshd brute
    force raised to level 12 on the demo lab". An auditor should know that this is a
    demo rule.
  ◦ Rule 5712, level 10, 2 hits: sshd brute force using non-existent users.
  ◦ Rule 533, level 7, 4 hits: listening ports changed.
  The report gives no MITRE ids and no agent ids per alert group, so I can't cite them
  from this report. (source: wazuh_posture_report)


What the report says it leaves out

• Compliance checks (SCA) are not in the signed report. The tool's reason: "no
  fleet-wide SCA summary read exists; generate with --agent to include SCA for one
  agent".
• The report also says it cannot establish:
  ◦ whether the Wazuh indexer is in sync with the manager;
  ◦ anything beyond the first 500 agents for coverage;
  ◦ the 6060 CVEs it did not rank.


Compliance checks (separate read, NOT covered by the signature)

All three agents score 41% on the CIS Ubuntu 24.04 benchmark v1.0.0: 99 checks pass, 138
fail and 42 are invalid, out of 279. The agents are 000 hyp-wazuh-01, 001 hyp-web-01
and 002 hyp-db-01. Their last scans ran on 2026-11-09 (source: wazuh_sca).

This is live data, but it is not part of the signed evidence. To get signed compliance
evidence, you would need a separate report generated per agent.

One mismatch to explain to the auditor: the report counts 2 active agents, but the live
agent list returns 3 (source: wazuh_agents). The extra one is the manager's own agent,
000. I believe the report leaves it out, but nothing returned states that, so treat it
as my reading of the gap.

Tools used: wazuh_posture_report, wazuh_agents, wazuh_sca

The signed report is  WPR_MW9YNNFH_98B591BF7D , and ARGUS made no changes to any server.

49.3s · 4 tools · 0 approvals

A real turn on our own install (CLI 0.5.0, 10 Nov 2026), replayed at speed. Names changed. Type the same words into quox.
✔ Machine verified against the CLI on 10 Nov 2026 · test report mw9ync

The answer gives the report id, its SHA-256, the WARD key id and public key, and the rule the hash is computed over: canonical JSON as in VOLT SPEC 3.2 (keys sorted at every level with the JavaScript default sort, undefined values dropped, arrays in order, no whitespace, every value written as JSON.stringify writes it, hashed as UTF-8 bytes). With that rule an auditor can check the report with no Quox software at all: rebuild the canonical JSON, compare its SHA-256 with report_sha256, confirm the same hash sits in the signed message, and verify the Ed25519 signature with any standard library and the key you sent them separately.

Each report also lists what it cannot establish, such as how many CVEs were ranked out of the total. Generating a report stores a signed record. It changes nothing on your servers.

stored reports
quox wazuh report listⓆ QUOX │ wazuh report list │ 11 reports ✔REPORT                   SCOPE      CREATED                   SHA256WPR_MW7XMPUM_F556DB709B  agent:002  2026-11-08T14:33:15.358Z  1a728925120f8df2WPR_MW7XMN6U_E954D11AB2  agent:001  2026-11-08T14:33:11.910Z  e7c3ee2f3bce956cWPR_MW7XMLVO_331572BDA7  agent:000  2026-11-08T14:33:10.212Z  8ffa91cd7caa6e15WPR_MW7XMAQ1_9140212E4E  fleet      2026-11-08T14:32:55.753Z  c9f437a874fbb214WPR_MW6B5GI6_38CD8E35FF  fleet      2026-11-07T11:16:12.366Z  251b7f16b988a317WPR_MW4INZPR_0FBAEDCABF  fleet      2026-11-06T05:11:02.031Z  5f53f19b492389a6
From a real run on our demo install.

quox wazuh report generate makes a new one; add --agent <id> for one server with its compliance (SCA) results.

2. Save the signed bundle and verify it offline

Section titled “2. Save the signed bundle and verify it offline”

This is the step the auditor runs on their own machine. It needs only the bundle and the public key.

offline verify
quox wazuh report show WPR_MW7XMAQ1_9140212E4E --out posture.json- vulnerability ranking covered 2002 of 8041 CVEs (top by CVSS plus every CISA KEV CVE)- which of your servers lack a Wazuh agent: no Quox-managed servers are registered to this organisation on this instance, so there is no server inventory to compare againstquox wazuh report verify posture.json --public-key 9pX7z2vnF5ufAMjQtke0VogdIL/2PcdpQ6NMljWcVaM=Ⓠ QUOX │ wazuh report verify │ verified (offline) ✔hash matches   binding ok   signature valid   key pinned (trust anchored)
From a real run on our demo install.

To show the check has teeth, we changed one number in the saved file (the count of KEV vulnerabilities, from 7 to 0) and verified it again.

tampered copy
quox wazuh report verify tampered.json --public-key 9pX7z2vnF5ufAMjQtke0VogdIL/2PcdpQ6NMljWcVaM=Ⓠ QUOX │ wazuh report verify │ NOT verified: report_hash_mismatch ✖hash DOES NOT match   binding FAILED   signature INVALID   key none (not trust anchored)exit 2
From a real run on our demo install.

Exit code 2 means the report does not verify. Exit 0 means the hash, the binding and the signature all check out against the pinned key.

Quox signed the report with the install’s WARD key and stored it for your organisation. Verification recomputes the hash and checks the signature locally, so the auditor does not have to trust your server, your network or you. The quox wazuh report verify command is the quick way; the canonical JSON rule in the answer is what lets an auditor repeat the same check with their own tools.