CVE-2026-21452 is a denial-of-service vulnerability classified under CWE-400 (Uncontrolled Resource Consumption) and CWE-789 (Uncontrolled Memory Allocation) affecting the MessagePack for Java (msgpack-java) serializer library prior to version 0.9.11. The vulnerability is triggered during the deserialization process of .msgpack files containing EXT32 extension objects. Although msgpack-java lazily parses extension headers, it trusts the declared payload length metadata when materializing extension data via the ExtensionValue.getData() method. An attacker can craft a syntactically valid .msgpack file with an EXT32 object specifying an excessively large payload length. When deserialized, the library attempts to allocate a byte array of this declared size without enforcing any upper bound, leading to unbounded heap memory allocation. This results in rapid JVM heap exhaustion, excessive garbage collection, or immediate termination with an OutOfMemoryError, causing denial of service. The attack requires no malformed bytes, no user interaction, and no privileges, making it remotely exploitable in environments that accept or fetch .msgpack artifacts, such as model registries, inference services, CI/CD pipelines, and cloud-based model hosting platforms. The malicious payload is extremely small yet valid, enabling it to evade basic validation and scanning mechanisms. The vulnerability is fixed in msgpack-java version 0.9.11.

For European organizations, this vulnerability poses a significant risk to the availability of services that rely on msgpack-java for deserializing .msgpack files, particularly in AI/ML model management, cloud services, and automated deployment pipelines. Exploitation can lead to JVM crashes and service outages, potentially causing cascading failures in production environments. This can disrupt critical business operations, degrade customer experience, and incur financial losses. Organizations using vulnerable versions in cloud-hosted or on-premises inference services may face remote denial-of-service attacks without warning. The small size and validity of the malicious payload complicate detection and prevention, increasing the likelihood of successful exploitation. Given the growing adoption of AI/ML workflows and cloud-native architectures in Europe, the impact could be widespread, especially in sectors like finance, healthcare, and manufacturing that depend on continuous service availability and model inference.

European organizations should immediately upgrade all msgpack-java dependencies to version 0.9.11 or later to eliminate this vulnerability. Additionally, implement strict input validation and filtering on all .msgpack files, especially those sourced externally or from untrusted origins. Employ runtime monitoring and JVM heap usage alerts to detect abnormal memory consumption patterns indicative of exploitation attempts. Integrate security scanning tools capable of analyzing .msgpack files for anomalous EXT32 payload lengths. Restrict network access and authentication for services that accept or fetch .msgpack artifacts to reduce exposure. Where feasible, sandbox or isolate deserialization processes to limit the impact of potential crashes. Finally, update incident response plans to include detection and mitigation steps for deserialization-based denial-of-service attacks.