Juju CVE-2026-32691

The CVSS 3.1 score of 5.3 reflects a Medium-severity vulnerability with specific characteristics: the attack vector is Network (AV:N), requiring an attacker to be on the same network or reachable via the Juju controller API; Attack Complexity is High (AC:H), indicating the race condition requires precise timing or repeated attempts; Privileges Required are Low (PR:L), as the attacker must be an authenticated unit agent; User Interaction is None (UI:N); Scope is Unchanged (S:U); and the impact is Confidentiality High (C:H) with no integrity or availability impact. This CVSS vector accurately reflects the real-world risk: while confidentiality impact is severe, exploitation demands both authentication and exploitation difficulty. The vulnerability is not listed in CISA's Known Exploited Vulnerabilities (KEV) catalog, and no public proof-of-concept code has been widely distributed, suggesting active exploitation in the wild is limited but plausible given that Juju is used in enterprise cloud environments. The EPSS score (if available) would likely be moderate, as the attack requires internal network access and unit agent credentials. The real priority depends on organizational exposure: environments with multi-tenant Juju deployments, untrusted unit agents, or sensitive secrets managed via Juju should prioritize patching immediately. Environments with strict network segmentation and trusted-only unit agents face lower practical risk but should still treat this as a medium-priority remediation.

An attacker with credentials for a malicious unit agent deployed in a Juju model observes or predicts that another unit is about to create a new secret (e.g., by monitoring model logs or API calls). The attacker quickly issues an ownership claim for that secret ID before the legitimate unit finishes initializing the secret's first revision. The Juju controller, due to the race condition, assigns ownership to the attacker's unit instead of the intended owner. The attacker can then read the secret's content, obtaining sensitive data such as database passwords, API keys, or encryption material. This scenario is most likely in multi-tenant or hostile environments where untrusted units share a model, or in scenarios where an attacker has compromised a unit's credentials.