Description

Improper caching of the original content type in Spring Cloud Stream Avro. Spring Cloud Stream 5.0.0 - 5.0.2 Spring Cloud Stream 4.3.0 - 4.3.3 Spring Cloud Stream 4.2.0 - 4.2.6

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "3.1",
  "pubDate": "2026-08-27T20:17:56.760Z",
  "pubdate": "2026-08-27T20:17:56.760Z",
  "executiveSummary": "A critical vulnerability exists within the Avro message handling logic of Spring Cloud Stream, where the original content-type is improperly cached. This defect allows for potential data misinterpretation or cross-protocol contamination within the messaging pipeline.\nThe vulnerability affects Spring Cloud Stream versions 5.0.0 through 5.0.2, 4.3.0 through 4.3.3, and 4.2.0 through 4.2.6. By failing to correctly isolate or invalidate the original content-type headers during the message processing lifecycle, the system may inadvertently treat incoming payloads with incorrect schema definitions or serialization formats.\nThe risk implication involves potential security bypasses or application crashes due to malformed data processing. An attacker with the ability to inject or manipulate message headers within the message broker infrastructure can exploit this caching behavior to influence how downstream consumers deserialize subsequent messages. No specific authentication or privilege requirements are noted beyond access to the message bus itself, making this a significant concern for distributed microservices architectures relying on Avro for data serialization.",
  "technicalDetails": "The root cause of this vulnerability lies in the state management of the message conversion pipeline within Spring Cloud Stream. When processing messages serialized via Apache Avro, the framework caches the original content-type header to optimize subsequent processing steps; however, this caching mechanism fails to account for message-to-message variance in a pooled or long-running consumer process.\nDuring the execution flow, when a message is received from the broker, the framework reads the Content-Type header to determine the appropriate MessageConverter. If the caching logic incorrectly persists a content-type from a previous message state into the current context, the converter may attempt to deserialize a payload using the wrong schema or serialization logic. Because Avro relies heavily on schema registry lookups and strict binary formatting, this mismatch often leads to deserialization errors, potentially causing an application-level denial-of-service or allowing an attacker to inject payloads that bypass schema validation if the cached metadata is sufficiently flexible.\nThe attack flow involves an adversary injecting a sequence of messages into the messaging infrastructure. First, the attacker sends a 'template' message that forces the consumer to cache a specific (potentially malformed or atypical) content-type. Subsequently, the attacker sends a primary target payload that the consumer will process under the influence of the stale cached metadata rather than the metadata provided within the actual message headers. This manipulation essentially tricks the deserialization engine into interpreting the malicious payload through an incorrect lens, which can result in unexpected application behavior or memory corruption depending on how the underlying Avro libraries handle the mismatched input stream.\nThis vulnerability is present within the message transformation and conversion components of Spring Cloud Stream. The flaw is not limited by network exposure beyond the reachability of the underlying message broker (e.g., Kafka or RabbitMQ). Since the vulnerability is embedded in the framework's core message handling logic, it affects all implementations that leverage the default Avro message converters provided by the vulnerable versions of the library."
}