Description

fluent-plugin-s3 is an Amazon S3 input and output plugin for Fluentd. From 0.7.0 to 1.8.4, the in_s3 input plugin reads the entire decompressed payload of gzip, lzma2, and lzop objects into memory without enforcing a decompression_size_limit. An attacker with permission to upload objects to the monitored S3 bucket can provide a highly compressed object that expands excessively when Fluentd processes it. The resulting memory exhaustion can cause the operating system to terminate the Fluentd process and disrupt all log collection on the affected node. This issue is fixed in version 1.8.5.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "2.7",
  "pubDate": "2026-09-14T18:17:47.513Z",
  "pubdate": "2026-09-14T18:17:47.513Z",
  "executiveSummary": "The fluent-plugin-s3 component in Fluentd is susceptible to a memory exhaustion vulnerability, classified as a Denial-of-Service (DoS) condition. This flaw exists within the in_s3 input plugin, which lacks a mandatory decompression_size_limit when processing compressed objects retrieved from Amazon S3.\nAn attacker possessing authorized write access to the monitored S3 bucket can craft a 'decompression bomb'—a highly compressed file designed to expand to a significantly larger size upon extraction. When Fluentd retrieves and processes this object, the plugin attempts to load the entire decompressed payload into system memory.\nThis behavior leads to uncontrolled memory consumption, potentially triggering the operating system's Out-Of-Memory (OOM) killer to terminate the Fluentd process. The impact is a total disruption of log collection and ingestion pipelines on the affected node, leading to visibility gaps and potential data loss. The vulnerability affects fluent-plugin-s3 versions 0.7.0 through 1.8.4. Upgrading to version 1.8.5 or later is the primary remediation strategy to enforce necessary resource constraints.",
  "technicalDetails": "The root cause of this vulnerability is the absence of an enforced memory buffer limit during the decompression of gzip, lzma2, and lzop streams within the in_s3 input plugin. When Fluentd triggers a fetch operation on an object stored in Amazon S3, the plugin initializes a stream reader to ingest the data. Because the plugin does not validate the final decompressed size against a predefined threshold, it blindly allocates memory to accommodate the expanding data stream.\nExploitation follows a specific attack flow: First, an attacker creates a maliciously crafted object where a small compressed file (e.g., a few kilobytes) represents a massive amount of data (e.g., several gigabytes) due to high compression ratios characteristic of gzip or similar algorithms. Second, the attacker uploads this object to the S3 bucket configured for monitoring by the vulnerable Fluentd instance. Third, the in_s3 plugin polls the bucket, detects the new object, and initiates a decompression routine. As the decompression engine expands the payload, the memory footprint of the Fluentd process increases linearly with the volume of the extracted data.\nThis leads to an immediate depletion of available system RAM. Once memory exhaustion thresholds are reached, the OS kernel, specifically the OOM killer, may identify the Fluentd process as the primary consumer of system resources and terminate it to preserve system stability. This results in a state of Denial-of-Service for the logging pipeline.\nThe vulnerability is inherent to the logic of the in_s3 input plugin for all versions from 0.7.0 up to 1.8.4. No complex authentication is required beyond the attacker's ability to successfully write objects to the target S3 bucket. The exploit is remote, as it relies on the interaction between the cloud storage source and the log aggregator. Post-exploitation, the impact remains restricted to the termination of the service; there is no documented evidence of remote code execution or privilege escalation, though the loss of log data can obscure further malicious activity or forensic analysis."
}