Technical impact
Kafka Streams computes the next punctuation time as the record timestamp plus a multiple of the punctuation interval without checking for overflow. A record whose timestamp is close to Long.MAX_VALUE causes the next punctuation time to wrap to a negative value. The punctuation queue then repeatedly re-queues that negative time while holding a lock, blocking the StreamThread indefinitely. The effect is a complete halt in processing for the affected StreamThread and any application partitions that rely on it, but it does not impact Kafka brokers or the cluster overall.
Affected Systems
The vulnerability affects Apache Kafka Streams libraries prior to versions 4.2.2, 4.3.2, and 4.4.0. It is triggered by writing a single record with a timestamp near Long.MAX_VALUE to a source topic whose stream topology uses stream-time punctuation. Brokers running Apache Kafka older than version 4.0.0 accept such timestamps unless the log.message.timestamp.after.max.ms setting is tightened; newer brokers reject the value by default, limiting exploitation to older or misconfigured installations.
Risk and Exploitability
Because the exploit requires only write access to a source topic and no elevated privileges, the attack vector is straightforward. The EPSS score is less than 1%, and the CVE is not listed in the CISA KEV catalog, indicating that large‑scale exploitation has not been observed. The CVSS score of 9.1 denotes high severity, but the total loss of availability for the affected Streams application makes the risk significant, especially in environments that still use older Kafka clusters or have relaxed timestamp bounds.