CVE-2026-12249: Canonical ADSys Trust Store Poisoning via Plaintext HTTP Certificate Auto-Enrollment

Synopsis

Trust store poisoning is the vulnerability class here, affecting Canonical ADSys versions through v0.16.2 on Ubuntu LTS and current releases. During Active Directory Certificate Services auto-enrollment, ADSys fetches the CA certificate over plaintext HTTP rather than HTTPS, allowing an unauthenticated attacker positioned on the network path between the managed Ubuntu host and the AD CS server to substitute an arbitrary root CA certificate. Successful exploitation causes the host's system-wide trust store to accept attacker-controlled certificates, enabling persistent interception and decryption of all subsequent TLS connections made by any application trusting the OS certificate store. A patched-image rebuild at v0.16.3 (and the corresponding Ubuntu LTS point releases) is available on HarborGuard for affected environments.

HarborGuard Coverage

Exploit Conditions

• Network reachabilityRequired

  The attacker must be positioned on the network path between the Ubuntu host and the AD CS CA server to intercept the plaintext HTTP certificate request.

• AuthenticationNot required

  No credentials or account are required; the attack targets an unauthenticated HTTP request initiated by the ADSys client.

• Victim interactionNot required

  No user action is needed; certificate auto-enrollment runs automatically as a system-level policy operation.

• Attack complexityDetail

  Exploitation requires the attacker to achieve a man-in-the-middle position on the relevant network segment, introducing an environmental prerequisite, though no race condition or memory-layout dependency is involved.

Blast Radius

• The attacker injects an arbitrary root CA certificate into the host's system-wide trust store via update-ca-certificates.
• All TLS clients on the affected host that rely on the OS trust store accept rogue certificates signed by the attacker-controlled CA for any domain.
• Subsequent HTTPS traffic from applications, package managers, and service agents on the host is transparently decrypted and readable by the attacker.
• The attacker can also modify in-transit responses, injecting malicious payloads into software updates, API responses, or authentication flows.