NewApp Monitoring is liveSign up free
AgentbeamBeta
Open-source package vulnerability · PyPI

CVE-2026-46556: FlaskBB: SSRF in get_image_info() via unrestricted avatar URL

mediumCVSS 3.1 6.5GHSA-xq32-9g7q-7297

Summary A Server-Side Request Forgery (SSRF) vulnerability in getimageinfo() allows any authenticated user to force the server to send HTTP requests to arbitrary internal endpoints, including cloud metadata services (e.g., AWS 169.254.169.254). This is a blind SSRF with confirmed internal port scanning and internal API triggering capabilities. CVSS 6.5 Medium.

Published May 21, 2026

Affected packages

flaskbb

Ecosystem
PyPI
Affected
<= 2.2.0
Fixed in
No patched version yet
Check
pip show flaskbb

Advisory details

From the GitHub Advisory Database (CC BY 4.0).

###Summary A Server-Side Request Forgery (SSRF) vulnerability in get_image_info() allows any authenticated user to force the server to send HTTP requests to arbitrary internal endpoints, including cloud metadata services (e.g., AWS 169.254.169.254). This is a blind SSRF with confirmed internal port scanning and internal API triggering capabilities. CVSS 6.5 Medium.

###Details In flaskbb/utils/helpers.py (line 571), the url parameter is passed directly to requests.get(url, stream=True) without any validation of scheme, host, or IP address.

python# flaskbb/utils/helpers.py:571
def get_image_info(url: str):
    r = requests.get(url, timeout=(3.05, 27), stream=True)

Attack chain:

POST /user/settings/user-details (avatar URL)
→ ValidateAvatarURL.validate()    # validators.py:103
→ check_image(avatar)             # helpers.py:628
→ get_image_info(url)             # helpers.py:571
→ requests.get(url)               # No domain/IP restriction
Entry points:

/user/settings/user-details (any authenticated user)
/admin/users/<id>/edit (admin only)

###PoC submit.zip

Log in to FlaskBB as any user Navigate to Settings → User Details Enter http://169.254.169.254/latest/meta-data/ as the avatar URL Submit the form The server sends a GET request to the internal metadata endpoint

Three exploitation channels confirmed:

Server-side request: Captured on mock metadata server Internal port scan: check_image() returns distinct errors (CONN_REFUSED, NO_CONTENT_LENGTH, TYPE_NOT_ALLOWED, SUCCESS) that map internal network topology Internal API triggering: Mock APIs on 127.0.0.1:9200 triggered via SSRF (deploy, shutdown, key dump endpoints)

###Impact Any authenticated user is impacted. Attackers can force the server to request internal services, cloud metadata endpoints, or private network resources. On cloud deployments (AWS/GCP/Azure), IAM credentials can be leaked. In production, any GET-triggered internal service is reachable: CI/CD webhooks, Elasticsearch, etcd, Consul, etc.

Credited to woohyunchoi-kentech, programsurf, yoonsh.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. No analysis has been written for this entry yet. Methodology