Description
SGLang through 0.5.21 contains a decompression amplification vulnerability in load_audio that allows remote attackers to exhaust memory by submitting highly compressed audio decoded without duration or sample limits. Attackers can supply small FLAC-encoded silence files through audio_url or audio_data inputs that expand into massive PCM buffers, exhausting worker memory and CPU for denial of service.
Published: 2026-10-12
Score: 8.7 High
EPSS: < 1% Very Low
KEV: No
Impact: Denial of Service
Action: Patch Now
AI Analysis

Impact

SGLang versions up to 0.5.21 contain a decompression amplification flaw in the load_audio routine that expands tiny, highly compressed FLAC pods of silence into enormous PCM buffers. By submitting these files, a remote attacker can cause the worker process to consume large amounts of memory and CPU, disrupting service availability. The vulnerability is a classic resource exhaustion attack with no authentication requirements.

Affected Systems

The flaw affects all releases of SGLang up to and including 0.5.21. Any installation that exposes the load_audio function via audio_url or audio_data inputs is vulnerable.

Risk and Exploitability

The CVSS score of 8.7 categorizes this issue as high severity. The EPSS score is not available, and the vulnerability is not listed in CISA’s KEV catalog, but the lack of limits on input size and sample count gives attackers a straightforward remote attack path that can lead to denial of service. Exploitation requires only the ability to submit arbitrary audio data to the vulnerable endpoint.

Generated by OpenCVE AI on October 12, 2026 at 15:36 UTC.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Upgrade to a version of SGLang released after 0.5.21 that contains the official fix for load_audio.
  • If an update is not feasible, apply input validation limits on audio_url and audio_data to constrain file size, parsed duration, and sample count before decoding.
  • Configure resource quotas or cgroup limits on the worker processes handling load_audio to cap memory and CPU usage, preventing a single request from exhausting system resources.

Generated by OpenCVE AI on October 12, 2026 at 15:36 UTC.

Tracking

Sign in to view the affected projects.

Advisories

No advisories yet.

History

Tue, 13 Oct 2026 19:30:00 +0000

Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'yes', 'Exploitation': 'poc', 'Technical Impact': 'partial'}, 'version': '2.0.3'}


Mon, 12 Oct 2026 19:45:00 +0000

Type Values Removed Values Added
First Time appeared Sgl-project
Sgl-project sglang
Vendors & Products Sgl-project
Sgl-project sglang

Mon, 12 Oct 2026 14:15:00 +0000

Type Values Removed Values Added
Description SGLang through 0.5.21 contains a decompression amplification vulnerability in load_audio that allows remote attackers to exhaust memory by submitting highly compressed audio decoded without duration or sample limits. Attackers can supply small FLAC-encoded silence files through audio_url or audio_data inputs that expand into massive PCM buffers, exhausting worker memory and CPU for denial of service.
Title SGLang through 0.5.21 Audio Decompression Amplification via load_audio
First Time appeared Lmsys
Lmsys sglang
Weaknesses CWE-409
CPEs cpe:2.3:a:lmsys:sglang:*:*:*:*:*:*:*:*
Vendors & Products Lmsys
Lmsys sglang
References
Metrics cvssV3_1

{'score': 7.5, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H'}

cvssV4_0

{'score': 8.7, 'vector': 'CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N'}


cve-icon MITRE

Status: PUBLISHED

Assigner: VulnCheck

Published:

Updated: 2026-10-13T18:24:26.414Z

Reserved: 2026-10-12T13:46:32.638Z

Link: CVE-2026-109226

cve-icon Vulnrichment

Updated: 2026-10-13T18:24:18.667Z

cve-icon NVD

Status : Deferred

Published: 2026-10-12T14:17:11.747

Modified: 2026-10-13T19:18:13.833

Link: CVE-2026-109226

cve-icon Redhat

No data.

cve-icon OpenCVE Enrichment

Updated: 2026-10-12T19:30:06Z

Weaknesses
  • CWE-409

    Improper Handling of Highly Compressed Data (Data Amplification)