Skip to main content

SGLang EUVDEUVD-2026-101757

| CVE-2026-110807 LOW
Improper Resource Shutdown or Release (CWE-404)
2026-10-14 cna@vuldb.com GHSA-9h5h-jqm8-rrcw
2.1
CVSS 4.0 · Vendor: vuldb
View

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain

  1. Access
    Reach SGLang generate endpoint
  2. Exploit
    Submit session_params with wrong-typed field
  3. Execution
    Trigger uncaught TypeError in SessionParams
  4. Impact
    Deny service to request

Vulnerability Assessment

Weakness
CWE
ASVS 5.0
V15 - Secure Coding and Architecture Resource release
Exploitation Conditions

Exploitation requires the ability to send requests to SGLang's 'generate' HTTP endpoint using the session feature, specifically supplying a session_params object in which one of the fields (id, rid, offset, replace, or drop_previous_output) is of an incorrect type - e.g., id or rid as a non-string, offset as a string or bool, or replace/drop_previous_output as a non-bool. In vulnerable versions (≤0.5.19) SessionParams(**obj.session_params) is constructed with no type validation, so the mismatched type raises an uncaught TypeError. Limiting factors: the CVSS assigns PR:L, implying some level of API access/privilege is needed, and no user interaction is required; impact is confined to availability of request handling (no data exposure or code execution). A key real-world caveat: SGLang generate endpoints are often deployed without authentication, which would make the condition trivially reachable (effectively PR:N) despite the vendor's PR:L rating.

Impact

This is a low-severity denial-of-service issue, and the available signals are consistent in pointing to limited real-world priority. The vendor-supplied CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L) yields a base score of only 2.1: network-reachable and low-complexity, but requiring some privilege (PR:L), with no confidentiality or integrity impact and only a low availability impact (VA:L). Exploitation status: publicly available exploit code exists (a repro script is published as a GitHub gist and the CVSS carries E:P/Proof-of-Concept), but this is NOT active exploitation - it is not listed in CISA KEV and no EPSS score was supplied, so widespread or automated exploitation is not evidenced. The source-code evidence (PR #41535) confirms the root cause and the limited impact: supplying a wrong-typed session_params field raises an uncaught TypeError during request tokenization, failing the request rather than achieving code execution or persistent server compromise, which is consistent with VA:L rather than a full-service crash. Note a plausible discrepancy worth verifying: the vector asserts PR:L (authenticated), yet SGLang's HTTP generate endpoint is frequently deployed without authentication, in which case real-world exploitation could be effectively unauthenticated (PR:N) - defenders should verify whether their deployment exposes the endpoint openly. Bottom line: genuine but low-priority; patch in routine cycles, prioritize only if an SGLang generate endpoint is internet-exposed without auth.

Exploit ScenarioDerived

An attacker with access to the SGLang generate endpoint sends a crafted request whose session_params object contains a field of the wrong type (for example, 'id' as a list or 'offset' as a string). SGLang constructs the SessionParams struct without type checking, raising an uncaught TypeError that aborts request processing and denies service. A public proof-of-concept repro script demonstrating the crash is available.

Monitor Track and reassess. Why: LOW severity (CVSS 0.1-3.9).
No patch listed yet 0.3% exploitation probability (EPSS)

Get an e-mail when this CVE is exploited, patched or re-scored. Free account.

Alert me

AnalysisAIv1 – Oct 15, 2026

Denial of service in sgl-project SGLang up to 0.5.19 allows remote attackers to crash request handling by submitting a malformed session_params object to the generate endpoint, where a field such as id, rid, offset, replace, or drop_previous_output is supplied with the wrong data type. Because SessionParams is constructed without type validation in vulnerable builds, the mismatched type raises an uncaught TypeError that aborts the affected request path; no data exposure or code execution results. Publicly available exploit code exists (a reproduction script is linked in the references) and the vendor was reportedly notified via an issue but had not responded at the time of disclosure; the issue is low severity (CVSS 2.1/3.1 score of 2.1 with A:L, E:P) and affects availability only.

RemediationAI

Upstream fix available (PR/commit); released patched version not independently confirmed. The correction is merged as pull request https://github.com/sgl-project/sglang/pull/41535, which adds a __post_init__ validator to SessionParams in python/sglang/srt/managers/io_struct.py (raising TypeError on non-string id/rid, non-int offset, and non-bool replace/drop_previous_output) and wraps SessionParams construction in tokenizer_manager.py so a TypeError becomes a handled ValueError; because this points to a PR rather than a tagged release, operators should track the repository for the first release incorporating it and, until upgrading, apply the patch or an equivalent validation shim. Mitigating controls with trade-offs include: enforce authentication and authorization in front of the generate endpoint, since SGLang generate endpoints are frequently deployed without auth and this materially reduces exposure (requires configuration and may break unauthenticated clients); restrict network reachability to the SGLang HTTP service via firewall or reverse-proxy allowlists so only trusted clients can reach /generate (limits legitimate distributed callers); and validate or reject session_params at a proxy/gateway, blocking non-string id/rid, non-integer offset, and non-boolean replace/drop_previous_output (adds a layer of request inspection and possible maintenance overhead). None of these controls eliminate the underlying need for the code-level fix. Vendor issue context is available at https://github.com/sgl-project/sglang/issues/41467.

Affected VersionsENISA

sglang 0.5.18
sglang 0.5.15
sglang 0.5.2
sglang 0.5.3
sglang 0.5.10
sglang 0.5.19
sglang 0.5.17
sglang 0.5.6
sglang 0.5.13
sglang 0.5.16
sglang 0.5.7
sglang 0.5.14
sglang 0.5.4
sglang 0.5.5
sglang 0.5.11
sglang 0.5.12
sglang 0.5.0
sglang 0.5.9
sglang 0.5.8
sglang 0.5.1

Affected Packages

euvd sglang no fix

Source: OSV (Open Source Vulnerabilities)

Affects your stack

Affects 1 package in euvd

Checking this CVE against your own stacks and servers is part of the Basic plan and up.

Technical ContextAI · show details

SGLang is a Python-based LLM serving framework whose request pipeline parses user-supplied JSON into internal structures. The vulnerable component is the SessionParams class defined in python/sglang/srt/managers/io_struct.py, a msgspec.Struct used to carry session state for the generate endpoint. In versions ≤0.5.19, tokenizer_manager.py builds this object with a bare SessionParams(**obj.session_params) call and performs no validation on the field types; msgspec.Struct permits the assignment so the type error surfaces only later when id/rid are expected to be strings, offset an int, and replace/drop_previous_output booleans. A value of the wrong type therefore triggers an uncaught TypeError that propagates through the tokenizer manager, disrupting request handling. The assigned CWE-404 (Improper Resource Shutdown or Release) captures the resulting abort/failed cleanup rather than a memory-corruption or injection class. CVSS 4.0 rates the issue AV:N/AC:L/AT:N/PR:L/UI:N with only VA:L (availability, low) and no confidentiality or integrity impact.

Know when Python gets its next vulnerability

  • Watch products and get an email when something new is published
  • Check this CVE against your own stack and servers
  • Exploit and PoC references, threat intelligence from 22 sources

No credit card · 7-day full trial

Severity by source

Vendor (vuldb) PRIMARY 2.1 LOW
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:P
vuln.today AI 4.3 MEDIUM

Network-reachable generate endpoint, low complexity, no user interaction; PR:L follows the vendor assumption of API access, and only a limited availability impact (A:L) with no C/I impact.

3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L
4.0 AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N

Primary rating from Vendor (vuldb).

CVSS VectorVendor: vuldb

Attack Vector
Network
Attack Complexity
Low
Attack Requirements
None
Privileges Required
Low
User Interaction
None
Vulnerable system: Confidentiality
None
Vulnerable system: Integrity
None
Vulnerable system: Availability
Low
Subsequent system: Confidentiality
None
Subsequent system: Integrity
None
Subsequent system: Availability
None

DescriptionCVE.org

A weakness has been identified in sgl-project SGLang up to 0.5.19. Affected by this vulnerability is an unknown functionality of the file python/sglang/srt/managers/io_struct.py of the component generate Endpoint. This manipulation of the argument session_params causes denial of service. The attack is possible to be carried out remotely. The exploit has been made available to the public and could be used for attacks. The project was informed of the problem early through an issue report but has not responded yet.

Threat Intelligence Overview

EPSS Score
0.3%
Top 75% exploitable
Prevalence
Medium
popular open-source LLM serving framework, common in ML inference deployments
Data Quality
5 sources
Confidence: Medium
EUVD NVD CVE.org EPSS Vulnrichment GitHub VulDB

How this decision was reached

The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.

SignalValue
Severity LOW
CISA KEV No
Public POC No
EPSS 0.3%
Patch available No
Active threat No
Deciding rule
LOW severity (CVSS 0.1-3.9)
What every action level means
Emergency Immediate action required. Criteria: CRITICAL + in CISA KEV (actively exploited) without patch, OR CRITICAL + EPSS > 30% without patch, OR CRITICAL + active threat signals without patch.
Act Now Prioritize within 24-48h. Criteria: in KEV with patch available, OR CRITICAL + public POC, OR HIGH + EPSS > 10%, OR CRITICAL + active threat, OR any CRITICAL severity.
This Week Schedule remediation this week. Criteria: HIGH severity (CVSS 7.0-8.9). Patch within normal change window.
This Month Plan within 30 days. Criteria: MEDIUM severity (CVSS 4.0-6.9), lower exploitation probability. Include in regular patch cycle.
Monitor Track and reassess. Criteria: LOW severity (CVSS 0.1-3.9), minimal exploitation risk. No immediate action needed.

SSVC AssessmentENISA

SSVC Decision· derived
Defer
Exploitation
Poc
System Exposure· derived
Open
Automatable
No
Human Impact· derived
Low
Technical Impact
Partial

Priority Score

10
Low Medium High Critical
KEV: 0
EPSS: 0
CVSS: +10
PoC: 0
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

Denial of Service attacks render applications or systems unavailable by overwhelming resources or triggering failure conditions.

More in Python

View all
CVE-2025-24016 CRITICAL POC
9.9 Feb 10

Wazuh SIEM platform versions 4.4.0 through 4.9.0 contain an unsafe deserialization vulnerability in the DistributedAPI t

CVE-2025-27520 CRITICAL POC
9.8 Apr 04

BentoML version 1.4.2 and earlier contains an unauthenticated remote code execution vulnerability through insecure deser

CVE-2025-2945 CRITICAL POC
9.9 Apr 03

pgAdmin 4 contains critical remote code execution vulnerabilities in the Query Tool download and Cloud Deployment endpoi

CVE-2013-5093 MEDIUM POC
6.8 Sep 27

The renderLocalView function in render/views.py in graphite-web in Graphite 0.9.5 through 0.9.10 uses the pickle Python

CVE-2025-32375 CRITICAL POC
9.8 Apr 09

BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Rated critica

CVE-2014-0224 HIGH POC
7.4 Jun 05

OpenSSL before 0.9.8za, 1.0.0 before 1.0.0m, and 1.0.1 before 1.0.1h does not properly restrict processing of ChangeCiph

CVE-2024-21644 HIGH POC
7.5 Jan 08

pyLoad download manager version prior to 0.5.0b3.dev77 exposes the Flask SECRET_KEY through an unauthenticated endpoint.

CVE-2026-33017 CRITICAL POC
9.3 Mar 17

Langflow (a visual LLM pipeline builder) contains a critical unauthenticated code execution vulnerability (CVE-2026-3301

CVE-2017-9462 HIGH POC
8.8 Jun 06

In Mercurial before 4.1.3, "hg serve --stdio" allows remote authenticated users to launch the Python debugger, and conse

CVE-2026-49869 CRITICAL POC
10.0 Jun 26

Unauthenticated remote code execution affects Kestra OSS (the open-source event-driven orchestration platform) prior to

CVE-2026-39987 CRITICAL POC
9.3 Apr 08

Unauthenticated remote code execution in Marimo ≤0.20.4 allows attackers to execute arbitrary system commands via the `/

CVE-2024-21645 MEDIUM POC
5.3 Jan 08

pyLoad is the free and open-source Download Manager written in pure Python. Rated medium severity (CVSS 5.3), this vulne

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy