CVE-2026-54475: 
Apache ActiveMQ Classic vulnerability analysis and mitigation

CVE-2026-54475 is a Missing Authorization vulnerability in Apache ActiveMQ Classic (Broker, All, and core packages) that allows an unauthenticated network attacker to consume messages from temporary destinations created by other connections. Temporary destinations in ActiveMQ Classic are designed to be isolated to the connection that created them; however, this isolation is enforced only on the client side, enabling any other connection to bypass the restriction and read messages intended for a different connection. The vulnerability was disclosed on June 29, 2026, by researcher Leon Johnson (GitHub: lokerxx) and affects Apache ActiveMQ before 5.19.8 and versions 6.0.0 through before 6.2.7. It carries a CVSS v3.1 base score of 7.5 (High) (Github Advisory, Openwall OSS-Sec).

The root cause is classified as CWE-862 (Missing Authorization) and CWE-1220 (Insufficient Granularity of Access Control). In Apache ActiveMQ Classic, temporary destinations (e.g., temporary queues or topics) are expected to be scoped to the connection that created them, but the broker does not enforce this isolation server-side — the check exists only in the client library. An attacker who establishes a separate broker connection can directly subscribe to or consume from another connection's temporary destination without any authorization check on the broker, effectively bypassing the intended access control boundary (Openwall OSS-Sec, Github Advisory). No authentication is required, and the attack complexity is low, as the broker is network-accessible and the bypass requires only a standard AMQP/OpenWire client connection (Github Advisory).

Successful exploitation allows an unauthenticated attacker to intercept and consume messages from temporary destinations belonging to other active connections, resulting in a high integrity impact — message data intended for a specific consumer can be silently stolen or tampered with. While confidentiality impact is not scored (the attacker reads messages rather than persistent data stores), sensitive application-layer data such as request/reply payloads, session tokens, or business logic messages transiting through temporary queues may be exposed. Availability is not directly impacted, but unauthorized consumption of messages can disrupt application workflows that rely on temporary destination isolation for request-reply patterns (Github Advisory, Red Hat Bugzilla).

1. Reconnaissance: Identify internet-facing or network-accessible Apache ActiveMQ broker instances running versions prior to 5.19.8 or between 6.0.0 and 6.2.6 using network scanning tools (e.g., Shodan, Censys, or Nmap targeting default ActiveMQ ports 61616/TCP for OpenWire or 5672/TCP for AMQP).
2. Establish a broker connection: Using a standard ActiveMQ client library (e.g., Apache ActiveMQ client for Java, Python's stomp.py, or any OpenWire-compatible client), establish a legitimate connection to the target broker without requiring credentials if the broker allows unauthenticated access, or with any valid credentials if authentication is enabled.
3. Enumerate temporary destinations: Monitor broker activity or use broker management APIs (e.g., JMX, ActiveMQ Web Console if accessible) to identify active temporary queue or topic names created by other connections (e.g., names following the pattern ActiveMQ.Advisory.TempQueue or similar).
4. Subscribe to the target temporary destination: Directly create a consumer/subscriber on the identified temporary destination name from the attacker-controlled connection. Because the broker does not enforce ownership checks server-side, the subscription succeeds.
5. Intercept messages: Receive and read messages intended for the legitimate owner connection, potentially capturing sensitive application data such as request-reply payloads, session tokens, or business logic content (Openwall OSS-Sec, Github Advisory).

• Network: Unexpected or anomalous client connections to the ActiveMQ broker (ports 61616, 5672, 61613) from unknown or unauthorized IP addresses; multiple simultaneous connections from different sources subscribing to the same temporary destination names.
• Logs: ActiveMQ broker logs showing consumer registrations on temporary destination names (e.g., temp-queue:// or temp-topic://) from connection IDs that did not create those destinations; unusual patterns of message consumption where messages are consumed faster than expected or by unexpected client IDs.
• Broker Management: JMX or Web Console showing consumers on temporary queues/topics associated with connection IDs that differ from the creating connection; unexpected subscriber counts greater than 1 on temporary destinations.
• Application: Legitimate consumers reporting missing or lost reply messages in request-reply messaging patterns; application-level errors indicating expected responses were never received on temporary reply queues (Openwall OSS-Sec, Red Hat Bugzilla).

The primary remediation is to upgrade Apache ActiveMQ to version 5.19.8 (for the 5.x branch) or 6.2.7 (for the 6.x branch), which enforce temporary destination isolation server-side on the broker (Openwall OSS-Sec, Github Advisory). No official configuration-based workaround has been published; therefore, upgrading is the only confirmed fix. As an interim measure, organizations should restrict network access to ActiveMQ broker ports using firewalls or network segmentation, enforce authentication on broker connections, and monitor for anomalous consumer activity on temporary destinations.

The vulnerability received coverage from several security news outlets including CyberSecurityNews, SecurityOnline, VPNCentral, and The Hacker News weekly recap, which grouped it alongside other Apache ActiveMQ vulnerabilities disclosed in the same advisory cycle (CyberSecurityNews). Red Hat tracked the issue via Bugzilla (Bug 2494840) with high severity, indicating potential downstream impact on Red Hat products shipping ActiveMQ (Red Hat Bugzilla). Community reaction has been moderate, with the vulnerability noted as automatable and network-accessible but tempered by the absence of a public PoC or active exploitation evidence.