Kruize Autotune EUVD-2026-109109

Exploitation requires network reachability to the Autotune service and that at least one monitoring datasource has been configured with authentication secrets (basic-auth password, bearer token, API key, or OAuth2 client secret) - if no secured datasources are configured there are no credentials to harvest. No authentication is needed because the REST API ships without any authentication layer (PR:N), and no user interaction is required. The principal limiting factor is the Adjacent attack vector (AV:A): the attacker must be able to reach the Autotune endpoints, which in typical deployments means positioning inside the same cluster or network segment rather than across the open internet. The vulnerable endpoints are specifically GET /datasources (list-datasources) and the datasource-metadata endpoint.

This is a credible, confirmed credential-exposure flaw rather than a hypothetical high-CVSS issue, but its reach is constrained. The CVSS 3.1 vector (AV:A/AC:L/PR:N/UI:N/S:C/C:H/I:N/A:N, 7.4) shows a low-complexity, unauthenticated path - PR:N confirms no authentication is required because the REST API ships with no authentication layer - with a scope change (S:C) justified by the fact that the leaked secrets grant access to a separate security authority (the backing monitoring stack). Impact is confidentiality-only (C:H, I:N, A:N): the flaw discloses stored credentials but does not itself modify or deny service. The attack vector is Adjacent (AV:A), not full Network (AV:N), which materially limits exposure - an attacker needs reachability to the Autotune service (typically cluster-internal), not arbitrary internet access. No EPSS score, no CISA KEV listing, and no public exploit were provided in the input, so there is no signal of active or probable exploitation; the GitHub reference is the vendor fix (PR 2038), not exploit code. The vuln.today priority is 'this_week' and the SSVC-style decision label aligns with a notable-but-not-emergency posture. Net: real and worth remediating promptly because exploitation is trivial once an attacker is adjacent and secrets are exposed in cleartext, but the adjacent-only vector and absence of exploitation evidence keep it below drop-everything urgency.

An attacker who has gained a foothold on the network adjacent to a Kruize Autotune deployment (for example a compromised pod or host in the same cluster segment) sends an unauthenticated GET request to the /datasources endpoint or the datasource-metadata endpoint. Because the datasource objects are serialized with a default Gson configuration and no field-exclusion strategy, the response returns the stored authentication secrets (basic-auth passwords, bearer tokens, API keys, OAuth2 client secrets) in cleartext, which the attacker then reuses to authenticate to the backing monitoring stack. No public exploit code was identified at time of analysis; exploitation requires only a standard HTTP client.