# Wazuh Manager 5.0 inventory_sync NDJSON Injection in OpenSearch _bulk API (GHSA-ff9g-85jq-r3g3, CVSS 10.0)

> An unauthenticated NDJSON injection flaw in the Wazuh Manager 5.0 inventory_sync subsystem lets any enrolled agent smuggle arbitrary OpenSearch _bulk operations through the unsanitized DataValue.index flatbuffer field, enabling deletion of security alerts, tampering of vulnerability/inventory data, and persistent payload injection into Kibana saved objects. Tracked as GHSA-ff9g-85jq-r3g3 with a maximum CVSS 3.1 score of 10.0.

- **Published:** 2026-06-15T00:00:00Z
- **Last reviewed:** 2026-06-15T00:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-0800
- **Card image:** https://intel.threadlinqs.com/og/threat/threat-intel-wazuh-inventory-sync-injection-tl-2026-0800.webp?v=a1cedc4f (Threadlinqs threat intel: GHSA-ff9g-85jq-r3g3 Wazuh NDJSON Flaw: Detection & IOCs (TL-2026-0800) — Critical severity, 17 IOCs, 9 detection rules.)
- **ID:** TL-2026-0800
- **Severity:** CRITICAL (CVSS 10)
- **Category:** VULNERABILITY
- **Status:** ACTIVE
- **Detections:** 9 (1 shown free below) · **IOCs:** 17

## Description

GHSA-ff9g-85jq-r3g3 is a critical (CVSS 10.0) injection vulnerability in the inventory_sync subsystem introduced in Wazuh Manager 5.0.0-beta1 — a code path that does not exist in the Wazuh 4.x branch. When an enrolled agent sends inventory data to the manager, the agent-controlled DataValue.index flatbuffer field is appended directly into the body of an OpenSearch _bulk NDJSON (newline-delimited JSON) request without any escaping or validation. In indexerConnectorSyncImpl.hpp the _index value is appended raw at both the bulkIndex and bulkDelete call sites (m_bulkData.append(index); // Untrusted input), while the adjacent _id field is correctly sanitized via appendEscapedId() — an asymmetry that is the root cause.

By embedding newline characters and crafted JSON action/metadata lines into the index field, an attacker injects additional bulk operations (delete, index, update) into the NDJSON payload the manager posts to the indexer. Because the injected operations execute under the manager's OpenSearch credentials — which in default deployments resolve to the admin account with the all_access role stored in the Wazuh keystore — a single agent-side injection is elevated to effective cluster-admin control over all indexed security data.

Exploitation is unauthenticated and network-reachable: Wazuh's default wazuh-authd configuration permits anonymous agent enrollment (use_password=no, ssl_verify_host=no) over TCP/1515, after which the attacker communicates over the standard remoted AES channel on TCP/1514. End-to-end public proof-of-concept code in the advisory (18-cve-f-remoted-aes.py) performs anonymous enrollment, AES-256-CBC session-key derivation, construction of a poisoned DataValue flatbuffer, and confirms targeted document deletion within ten seconds; a second harness (15-cve-f-sink-harness.cpp) isolates the NDJSON injection at the bulkIndex sink.

Impact spans defense evasion and data integrity destruction across an entire SIEM deployment: attackers can delete alerts in wazuh-alerts-* indices to erase post-compromise evidence, overwrite inventory and vulnerability state for monitored agents, write persistent payloads into .kibana_1 saved objects, and perform cross-tenant manipulation in shared environments. The Wazuh security team shipped three layered fixes in 5.0.0-beta3: sink-level escaping of _index (mirroring the existing _id escaping), boundary validation of index names in inventorySyncFacade.hpp, and downscoping the keystore credential from admin/all_access to a narrower wazuh-server role. No CVE identifier was assigned at publication; the issue is tracked solely via GHSA-ff9g-85jq-r3g3.

## MITRE ATT&CK

- T1595 Active Scanning
- T1046 Network Service Discovery
- T1190 Exploit Public-Facing Application
- T1133 External Remote Services
- T1078.001 Default Accounts
- T1552.001 Credentials In Files
- T1685 Disable or Modify Tools
- T1070 Indicator Removal
- T1685.006 Clear Linux or Mac System Logs
- T1505 Server Software Component
- T1213 Data from Information Repositories
- T1485 Data Destruction
- T1565.001 Stored Data Manipulation
- T1565.002 Transmitted Data Manipulation
- T1491.001 Internal Defacement

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**Bulk Delete / Drop in wazuh-alerts-* Doc Count From Indexer Service Account (Evidence Destruction)** (`TL-2026-0800-SPL-003`)

Detects the anti-forensic impact stage of GHSA-ff9g-85jq-r3g3: bulk delete operations against wazuh-alerts-* (or other security indices) executed under the wazuh-indexer/admin all_access keystore credential, and the corresponding sudden drop in alert document counts inconsistent with normal ingest/retention. This is defense evasion and data destruction used to erase post-compromise evidence.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1485, T1070, T1685, T1565.001
- **Data sources:** opensearch_audit, wazuh_indexer
- **Severity:** critical
- **Confidence:** high

### Splunk SPL

```spl
index=wazuh_indexer sourcetype="opensearch:audit" action="indices:data/write/bulk" request="*\"delete\"*"
| rex field=request "_index\"\s*:\s*\"(?<target_index>wazuh-[^\"]+)\""
| search target_index="wazuh-alerts*" OR target_index="wazuh-states-vulnerabilities*"
| stats count as delete_ops dc(target_index) as indices values(target_index) as targets min(_time) as firstSeen max(_time) as lastSeen by principal src_ip
| where delete_ops>0 AND (principal="admin" OR principal="*all_access*" OR principal="wazuh-indexer")
| eval note="bulk delete on security indices under indexer service account"
| `ctime(firstSeen)` | `ctime(lastSeen)`
```

### Microsoft KQL

```kql
let lookback = 24h;
OpenSearchAudit_CL
| where TimeGenerated > ago(lookback)
| where Action_s == "indices:data/write/bulk" and Request_s has '"delete"'
| extend TargetIndex = extract(@"_index\"\s*:\s*\"(wazuh-[^\"]+)\"", 1, Request_s)
| where TargetIndex startswith "wazuh-alerts" or TargetIndex startswith "wazuh-states-vulnerabilities"
| where Principal_s in ("admin", "wazuh-indexer") or Principal_s has "all_access"
| summarize DeleteOps = count(), Targets = make_set(TargetIndex), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated) by Principal_s, SrcIp_s
| where DeleteOps > 0
| order by DeleteOps desc
```

### Sigma

```yaml
title: Bulk Delete Against wazuh-alerts-* From Wazuh Indexer Service Account (Evidence Destruction)
id: b2e1d1ef-0800-4c11-8d22-3f5a7b9c1e33
status: experimental
description: Detects bulk delete operations against wazuh-alerts-* or vulnerability-state indices executed under the admin/all_access or wazuh-indexer service account, the data-destruction / indicator-removal impact of GHSA-ff9g-85jq-r3g3 used to erase forensic evidence.
references:
    - https://github.com/wazuh/wazuh/security/advisories/GHSA-ff9g-85jq-r3g3
    - https://attack.mitre.org/techniques/T1485/
    - https://attack.mitre.org/techniques/T1070/
author: Threadlinqs Team
date: 2026-06-15
tags:
    - attack.impact
    - attack.t1485
    - attack.t1565.001
    - attack.defense_evasion
    - attack.t1070
    - attack.t1685
logsource:
    product: opensearch
    category: audit
detection:
    selection:
        action: 'indices:data/write/bulk'
        request|contains: '"delete"'
        index|startswith: 'wazuh-alerts'
    filter_legitimate:
        layer: 'ism'
    condition: selection and not filter_legitimate
falsepositives:
    - ISM/ILM retention rollover or snapshot lifecycle deletes during scheduled maintenance
    - Authorized administrative index cleanup
level: high
```

**Known false positives:**

- ISM/ILM retention rollover or snapshot lifecycle deletes during scheduled maintenance windows
- Authorized administrative index cleanup or reindex operations

**Tuning notes:** Exclude the ISM/ILM service principal and known retention schedules; pair with a metric monitor that alerts on a >X% drop in wazuh-alerts-* doc count over a short interval. Downscope the indexer keystore credential so any all_access delete becomes inherently anomalous.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- OpenSearch _bulk NDJSON Action/Source Line Ratio Anomaly (inventory_sync Injection) (locked: unlock with TL Creds)
- Unattended Write to [indicator] Saved Objects Under Indexer Credential (Persistence) (locked: unlock with TL Creds)
- Privileged all_access Keystore Credential Used for Destructive Bulk Ops (Credential Abuse / Impact) (locked: unlock with TL Creds)
- Newline/CRLF Bytes In Wazuh inventory_sync Index Field (Injection Primitive) (locked: unlock with TL Creds)
- Wazuh Keystore Credential Read / indexer Service Account Spike (Credentials-in-Files Access) (locked: unlock with TL Creds)
- Wazuh Alert Index Tampering / Mass Document Drop (Indicator Removal & Stored-Data Manipulation) (locked: unlock with TL Creds)
- Wazuh inventory_sync NDJSON Injection — Smuggled _bulk Action via Unescaped _index (Exploitation) (locked: unlock with TL Creds)
- Anonymous Wazuh Agent Enrollment Followed by Immediate inventory_sync Traffic (Exploitation Setup) (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sources

- [GitHub Security Advisory GHSA-ff9g-85jq-r3g3](https://github.com/wazuh/wazuh/security/advisories/GHSA-ff9g-85jq-r3g3)
- [GitHub Advisory Database entry GHSA-ff9g-85jq-r3g3](https://github.com/advisories/GHSA-ff9g-85jq-r3g3)
- [Wazuh Vulnerability Lets Attackers Inject Malicious OpenSearch Bulk Operations](https://cybersecuritynews.com/wazuh-vulnerability/)
- [Critical Wazuh Flaw Enables Alert Tampering and Evidence Deletion](https://cyberpress.org/critical-wazuh-flaw/)
- [Wazuh Releases (5.0.0 beta line)](https://github.com/wazuh/wazuh/releases)
- [CWE-74: Improper Neutralization of Special Elements in Output Used by a Downstream Component (Injection)](https://cwe.mitre.org/data/definitions/74.html)
- [CWE-93: Improper Neutralization of CRLF Sequences (CRLF Injection)](https://cwe.mitre.org/data/definitions/93.html)
- [CWE-863: Incorrect Authorization](https://cwe.mitre.org/data/definitions/863.html)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-0800
