VAITP Dataset

← Back to the dataset

CVE-2026-112082

Remote DoS via expert_distribution_handle in SGLang ≤0.5.19 (public exploit).

  • CVSS 7.7
  • CWE-404
  • Input Validation and Sanitization
  • Remote

A vulnerability was identified in sgl-project SGLang up to 0.5.19. This impacts the function Scheduler.expert_distribution_handle of the file python/sglang/srt/eplb/expert_distribution.py of the component Expert-Distribution Recording Endpoint. Such manipulation leads to denial of service. It is possible to launch the attack remotely. The exploit is publicly available and might be used. The pull request to fix this issue awaits acceptance.

CVSS base score
7.7
Published
2026-10-19
OWASP
A06 Vulnerable and Outdated Components
Orthogonal defect classification
Timing/Serialization
Code defect classification
Timing Issues
Category
Input Validation and Sanitization
Subcategory
Insecure Parsing or Deserialization
Accessibility scope
Remote
Impact
Denial of Service (DoS)
Affected component
Python
Fixed by upgrading
Yes

Solution

Upgrade SGLang to version 0.5.20, which contains the fix for the Scheduler.expert_distribution_handle DoS vulnerability.

Vulnerable code sample

import json

class Scheduler:
    def expert_distribution_handle(self, data_str):
        # VULNERABLE: uses eval on untrusted input leading to DoS or code exec
        config = eval(data_str)  # noqa: eval-used
        # process config (placeholder)
        return config.get("distribution", {})

Patched code sample

import json

class Scheduler:
    def expert_distribution_handle(self, data_str):
        # FIX: replace eval with safe JSON parsing
        config = json.loads(data_str)
        # process config (placeholder)
        return config.get("distribution", {})

Published fix

Public sources link this CVE to the changes below. The code samples above were generated to illustrate the defect; these links lead to the code the project actually changed.

  • Reject disabled expert recorder controls before scheduler dispatch (sgl-project/sglang)
    python/sglang/srt/entrypoints/http_server.py · dump_expert_distribution_record_async, python/sglang/srt/entrypoints/http_server.py · start_expert_distribution_record_async, python/sglang/srt/entrypoints/http_server.py · stop_expert_distribution_record_async, python/sglang/srt/managers/tokenizer_control_mixin.py, python/sglang/srt/managers/tokenizer_control_mixin.py · TokenizerControlMixin.dump_expert_distribution_record, python/sglang/srt/managers/tokenizer_control_mixin.py · TokenizerControlMixin.start_expert_distribution_record, python/sglang/srt/managers/tokenizer_control_mixin.py · TokenizerControlMixin.stop_expert_distribution_record

Cite this entry

@misc{vaitp:cve2026112082,
  title        = {{Remote DoS via expert_distribution_handle in SGLang ≤0.5.19 (public exploit).}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2026},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2026-112082},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2026-112082/}}
}

Download this entry as JSON Download the whole dataset

Introducing the "VAITP dataset": a specialized repository of Python vulnerabilities and patches, meticulously compiled for the use of the security research community. As Python's prominence grows, understanding and addressing potential security vulnerabilities become crucial. Crafted by and for the cybersecurity community, this dataset offers a valuable resource for researchers, analysts, and developers to analyze and mitigate the security risks associated with Python. Through the comprehensive exploration of vulnerabilities and corresponding patches, the VAITP dataset fosters a safer and more resilient Python ecosystem, encouraging collaborative advancements in programming security.

The supreme art of war is to subdue the enemy without fighting.

Sun Tzu – “The Art of War”

:: Shaping the future through research and ingenuity ::

Before you use the VAITP AI models

Please read and accept the following disclaimer:

By choosing "I understand and accept", you acknowledge that you understand and agree to these terms.