Wazuh CVE-2026-67307

Exploitation requires control of a host already enrolled as a Wazuh 5.0.0-beta1 or beta2 agent - an external or unenrolled attacker cannot trigger this path without first obtaining an enrolled agent credential (PR:L). The attacker must be capable of crafting or replaying FlatBuffer inventory-sync Start messages with modified cluster metadata fields. For the most severe cross-cluster integrity impact (SI:H), a shared OpenSearch or Elasticsearch indexer must be in use across multiple independent Wazuh clusters, and numeric agent IDs must collide between the attacker's cluster and the target (AT:P). Single-cluster deployments limit exploitation to self-contained inventory record tampering within that cluster (VI:H only). High attack complexity (AC:H) further constrains exploitation to actors with knowledge of the Wazuh 5.x FlatBuffer protocol internals. Wazuh 4.x stable deployments are entirely unaffected.

The CVSS 4.0 score of 7.0 with vector AV:N/AC:H/AT:P/PR:L/UI:N/VC:N/VI:H/VA:N/SC:N/SI:H/SA:N reflects a genuine but structurally constrained risk. AC:H and AT:P together signal that exploitation requires non-trivial prerequisites: the attacker must control an already-enrolled Wazuh agent (PR:L) and the worst-case cross-cluster poisoning further demands a shared-indexer multi-cluster deployment with colliding numeric agent IDs (AT:P). The integrity impact is confirmed high on both the local cluster (VI:H) and subsequent cross-cluster records (SI:H) in that architecture; there is no confidentiality or availability impact. No EPSS score was provided in the source data, and the vulnerability is not listed in CISA KEV; no public exploit code has been identified. The most material real-world risk is in MSSP or enterprise environments running Wazuh 5.0.0-beta with multiple clusters sharing one indexer, where a rogue or compromised enrolled agent could poison a separate tenant's vulnerability scan records - for example, suppressing real findings or injecting false positives to degrade detection fidelity. Single-cluster deployments face only self-contained inventory tampering. Because the affected versions are pre-release betas (5.0.0-beta1 and beta2), production deployment footprint is inherently small, which substantially limits aggregate exposure despite the high per-instance integrity impact.

An attacker who controls or compromises a host already enrolled as a Wazuh agent modifies the agent binary or intercepts the agent communication channel to inject inventory-sync Start FlatBuffer messages with a forged cluster_name matching a co-tenant cluster in the same shared-indexer deployment. Where numeric agent IDs collide, the malicious messages cause the Wazuh manager to index vulnerability documents under the target cluster's namespace, allowing the attacker to suppress real CVE findings or inject phantom vulnerabilities into the target cluster's security dashboard. No public proof-of-concept exists, but the technique is derivable from knowledge of the FlatBuffer schema.