HAProxy Ingress Controller CVE-2025-59303

Exploitation requires two specific preconditions: (1) the HAProxy Kubernetes Ingress Controller must be running with the non-default 'config-snippets' feature flag enabled, and (2) the attacker must already possess Kubernetes RBAC create/update permissions on Ingress resources (PR:L). When both hold, submitting a config snippet causes the controller to return the ingress token secret in its response. Limiting factors that reduce real-world risk: the config-snippets flag is an opt-in feature not enabled by default, and the attacker needs a legitimate (or compromised) authenticated identity with write access to Ingress objects - anonymous or read-only accounts cannot exploit this. Affected versions are before 3.1.13; the issue is remediated in HAProxy Enterprise Kubernetes Ingress Controller 3.0.16-ee1, 1.11.13-ee1, and 1.9.15-ee1.

This is a moderate-priority, configuration-dependent information disclosure rather than a widespread emergency. The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:L/A:N, base 6.4) indicates a network-reachable flaw requiring low privileges and no user interaction, with a scope change (S:C) reflecting that the attacker crosses from their namespaced RBAC role into the controller's trust boundary to retrieve a secret it manages. Crucially, PR:L confirms the attacker must already hold create/update permissions on Ingress resources - this is an authenticated, privileged-insider or compromised-service-account scenario, not anonymous remote exploitation. The EPSS score is low (0.26%, 16th percentile), there is no CISA KEV listing, and no public exploit has been identified, all of which align: this is a recently disclosed, niche-precondition issue unlikely to see mass scanning. The single material caveat is the non-default precondition - exploitation only applies when the 'config-snippets' feature flag is enabled - which the CVSS base vector does not express and which materially lowers real-world exposure below what the 6.4 score alone suggests. No conflicting signals were found across sources; the chief limitation is the thinness of sourcing (a single vendor blog reference).

In a cluster where the HAProxy Kubernetes Ingress Controller is deployed with the config-snippets feature flag enabled, an attacker who already holds create/update permissions on Ingress objects (e.g., a tenant in a multi-tenant cluster or a compromised namespaced service account) submits a crafted Ingress with a malicious config snippet. The controller's handling of the snippet returns the ingress token secret in its response, exposing a credential the attacker's role should never see. No public proof-of-concept is currently available, and low attack complexity means the request itself is trivial once the privilege and configuration preconditions are met.