Get ahead
VMware offers training and certification to turbo-charge your progress.
Learn moreWhen mapping inbound record headers, JsonKafkaHeaderMapper and DefaultKafkaHeaderMapper (used by the default message converters) deserialize each header value into the type named for it in the producer-controlled spring_json_header_types header, provided that type's package is trusted. Because java.lang and java.util are trusted by default, a producer can declare a type with a very large number of array dimensions, such as java.lang.String followed by many [] suffixes, or an array of java.util.Timer.
Resolving the deeply nested array type recurses once per dimension and overflows the stack of the consumer thread with a StackOverflowError. Deserializing a java.util.Timer array starts a new thread for each element, so a large array can exhaust the native threads of the application. In both cases, the error is not caught by the mapper, so the listener container stops before the record is committed, and stops again on the same record after a restart. A malicious producer can use a single record to repeatedly stop consumption.
An application is affected if it consumes records from a topic that an attacker can produce to and maps headers with JsonKafkaHeaderMapper or DefaultKafkaHeaderMapper, which is the default for @KafkaListener methods and other listeners using the messaging message converters.
Spring for Apache Kafka:
| Fix version | Availability |
|---|---|
| 4.1.3 | OSS |
| 4.1.2.1 | Enterprise Support Only |
| 4.0.9 | OSS |
| 4.0.8.1 | Enterprise Support Only |
| 3.3.18 | Enterprise Support Only |
| 2.9.16 | Enterprise Support Only |
| 2.8.14 | Enterprise Support Only |
No further mitigation steps are necessary.
This vulnerability was discovered and responsibly reported by JIHUN KIM.
To report a security vulnerability for a project within the Spring portfolio, see the Security Policy