CVE-2026-105820 in Vault Enterprise
Summary
by MITRE • 10/08/2026
Vault's ACL policy cache allowed namespace traversal when policy names contained path traversal constructs. This may allow a token assigned specially crafted policy names to use the capabilities of policies defined in other namespaces, including the root namespace. This vulnerability (CVE-2026-105820) is fixed in Vault Enterprise 2.1.2, 1.21.12, 1.20.17, and 1.19.23. Vault Community Edition does not support namespaces, and is not affected.
Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.
Analysis
by VulDB Data Team • 10/08/2026
The vulnerability identified as CVE-2026-105820 represents a critical security flaw within the HashiCorp Vault Access Control List policy caching mechanism, specifically affecting environments that utilize namespace isolation features available in Vault Enterprise editions. The core technical issue stems from an insufficient validation of path traversal constructs embedded within ACL policy names during the cache lookup process. When a token is assigned policies with maliciously crafted names containing sequences designed to traverse directory boundaries, such as dot-dot-slash patterns or similar encoding variants, the caching subsystem fails to properly sanitize these inputs before resolving their location relative to the current namespace context. This failure allows an attacker who controls policy assignment for a specific entity or token to bypass the intended logical separation between namespaces. By exploiting this path traversal weakness in the cache resolution logic, it becomes possible to reference and inherit capabilities defined in policies located outside the authorized scope of the requesting principal's designated namespace.
The operational impact of this vulnerability is severe due to its potential for privilege escalation across administrative boundaries. In a multi-tenant or segmented deployment architecture, namespaces are typically used to isolate different teams, projects, or customer environments from one another, ensuring that access controls in one segment do not inadvertently grant permissions in another. An attacker leveraging CVE-2026-105820 can effectively escape this isolation boundary. Specifically, the vulnerability allows a token operating within a restricted child namespace to resolve policy references that point back to the root namespace or other sibling namespaces. Consequently, the attacker gains access to capabilities and secrets intended for higher privileged roles or different organizational units. This could lead to unauthorized reading of sensitive data, modification of critical configurations, or execution of administrative actions across multiple isolated environments, fundamentally undermining the security model provided by Vault's multi-tenancy features.
This flaw aligns with CWE-20 Improper Input Validation and CWE-78 OS Command Injection principles related to path traversal, as it involves failing to neutralize special elements within user-supplied input that could alter the intended execution flow or resource access path. In terms of offensive security frameworks, this vulnerability facilitates lateral movement and privilege escalation across logical boundaries, mapping closely to ATT&CK techniques involving unauthorized access to resources in adjacent domains or namespaces. The exploitation does not require direct code execution but rather manipulates the identity and permission resolution mechanisms inherent to the authentication and authorization subsystems.
To mitigate this risk, organizations running affected versions of Vault Enterprise must upgrade immediately to one of the patched releases: Vault Enterprise 2.1.2, 1.21.12, 1.20.17, or 1.19.23. It is important to note that Vault Community Edition does not support namespaces and therefore is not susceptible to this specific namespace traversal attack vector, although general security best practices should still be applied. For environments where an immediate upgrade is not feasible due to operational constraints, temporary mitigations may include restricting the ability of lower-privileged users or services to assign custom policies with complex naming conventions until patches can be deployed. Additionally, auditing logs for unusual policy assignment patterns involving special characters in policy names can help detect potential exploitation attempts. Regular review of ACL policies and enforcement of strict naming conventions that prohibit path traversal sequences are recommended defensive measures to reduce the attack surface associated with input validation flaws in configuration management systems.