Elastic Patches 14 Security Flaws, Including One Enabling Cross-Tenant Data Interception

Elasticsearch has an authorization-bypass flaw (CWE-639) in how it handles cross-cluster search requests under the Remote Cluster Security (RCS) 2.0 model. An authorization check validates the request against one identifying attribute of the target shard, while a separate, independently supplied attribute in the same request selects the shard that is actually accessed. A holder of a cross-cluster API key authorized for one index can point those two attributes at different indices, so the request is authorized against an allowed index while operating on a different, unauthorized index. That can disclose document contents, field mappings, and other metadata, and in limited cases can change retention-lease state on the unauthorized index. It is not listed in CISA KEV, and no public proof-of-concept is known.

Do: Apply the Elastic security update for this issue as soon as it is available for your release, and until then restrict who can hold cross-cluster API keys under Remote Cluster Security (RCS) 2.0. Review cross-cluster search API key privileges so each key is limited to the indices it should reach, and check access and retention-lease logs for requests whose shard identifiers do not match the index the key is authorized for.

CVE-2026-102406 is an authorization bypass through a user-controlled key (CWE-639) in Kibana's Fleet package installation process. A user who holds delegated Fleet package-management privileges, but not direct Elasticsearch administrative privileges, can claim a data stream identifier already in use by another tenant on the same Kibana deployment, and Fleet applies the uploaded package's generated index and ingest-pipeline settings without verifying ownership. The attacker can redirect that tenant's data stream through infrastructure they control, so subsequently ingested data can be disclosed or modified and does not reach its intended destination. It affects users or teams sharing one Kibana deployment, not separate Elastic Cloud organizations, and interception can continue after the malicious package is removed until the affected infrastructure is remediated separately. No public proof of concept is known, and the CVE is not listed in CISA's Known Exploited Vulnerabilities catalog.

Do: Apply the fixed Kibana release from Elastic's advisory for CVE-2026-102406 as soon as it is identified; no patched version range was included in the supplied data. Until then, limit who can hold delegated Fleet package-management privileges and audit data-stream identifiers, index settings, and ingest pipelines for claims or rewrites made by a package a delegated user installed. Removing the package is not enough: separately restore ownership and routing of affected data streams so ingestion returns to the intended destination, and review logs for disclosure or modification of another tenant's data.

CVE-2026-102413 is an uncaught exception (CWE-248) in Elastic Defend's Elastic Endpoint component that can crash the agent when it processes a specially crafted file name. The fault occurs during file-path handling on Windows when the system locale is one of certain configurations, including Chinese, Japanese, and Korean. A local attacker needs no privileges and no user interaction; the endpoint process crashes and restarts repeatedly, which can degrade or disable real-time malware prevention and behavioral detection on that host for as long as the condition persists. There is no confidentiality or integrity impact. It is not listed in CISA KEV, and no public proof-of-concept is known.

Do: Update Elastic Defend's Elastic Endpoint agent to the fixed release in Elastic's advisory for this CVE as soon as it is available, and confirm the agent stays running after the update. Until then, watch Windows hosts configured with Chinese, Japanese, or Korean locales for Elastic Endpoint crash-and-restart loops, and treat a looping agent as a loss of real-time malware prevention and behavioral detection on that host.