Skip to content

CVE-2026-112082: SGLang Expert Parallelism Load Balancing (EPLB) Worker Denial of Service

Density:
C-Level Executive Briefing CVE-2026-112082 HIGH SCORE 91/100

SGLang LLM serving engine through version 0.5.19 contains an unhandled exception vulnerability in Scheduler.expert_distribution_handle during Mixture-of-Experts (MoE) routing, allowing remote callers to trigger fatal worker process crashes.

HASS

HERMES THREAT SCORE

Confidence: 90%
0 / 100
MINIMAL

Measures specific systemic risk arising from autonomy, tool authority, and cascading execution.


SGLang LLM serving engine through version 0.5.19 contains an unhandled exception vulnerability in Scheduler.expert_distribution_handle during Mixture-of-Experts (MoE) routing, allowing remote callers to trigger fatal worker process crashes.


SGLang provides high-performance inference serving for open-weights Mixture-of-Experts (MoE) LLMs.

  • During dynamic Expert Parallelism Load Balancing (EPLB), Scheduler.expert_distribution_handle expects valid tensor dimension indices.
  • Malformed prompt token lengths with edge-case padding cause an index out-of-bounds error that is not caught by the scheduling loop.
  • The unhandled exception brings down the Ray/PyTorch worker process, terminating all concurrent inference sessions on the affected GPU node.
  • Fixed in 0.5.20 with boundary assertion checks and graceful fallback scheduling.

ProductVulnerable VersionsFixed ReleasePatch Availability
SGLang LLM EngineSGLang ≤ 0.5.190.5.20Vendor Patch Available

  1. Immediate Patch Deployment: Upgrade affected systems to 0.5.20.
  2. Action Item: Upgrade SGLang to version 0.5.20 or disable dynamic MoE expert rebalancing in inference worker configurations.
  3. Continuous Monitoring: Monitor endpoint and network telemetry for compromise indicators.
๐Ÿ•ธ๏ธ Connected Knowledge Graph & Provenance

CVE-2026-112082: SGLang Expert Parallelism Load Balancing (EPLB) Worker Denial of ServiceVULNERABILITY

Connected Nodes: 1
Active Relationships (Outgoing)
→ affectsPRODUCTSGLang Structured Generation & LLM Inference Engine
98% VERY_HIGH

Software platform affected by security vulnerabilities and agentic attack patterns.

๐Ÿ” Why is this related? (Evidence & Provenance)

“Confirmed security vulnerability in SGLang Structured Generation & LLM Inference Engine documented in Hermes dossier.”

Supporting Verified Evidence:
C-Level Executive Briefing HERMES-SEC-BRIEF HIGH
Generate Executive Briefing Sheet (Print / PDF)