Summary
Mistune: Denial of Service, RecursionError via Excessive Emphasis Markers in Markdown
Mistune v3.3.2 is vulnerable to a Denial of Service (DoS) attack via uncontrolled recursion in the HTML rendering of deeply-nested emphasis tokens. By submitting Markdown containing approximately 1,000 consecutive asterisk characters, an attacker causes the Python process to crash with RecursionError.
Details
The InlineParser's _process_emphasis_delimiters() creates deeply nested tokens from consecutive emphasis markers (every 2 asterisks add one nesting level). With 1,000 consecutive asterisks, approximately 500 levels of nesting are produced. The HTMLRenderer.render_token() method (src/mistune/renderers/html.py:40-57) renders these tokens recursively: when a token has children, line 48 calls self.render_tokens(token['children'], state), entering child rendering. Each nesting level produces ~2 stack frames, so 500 levels ≈ 1,000 frames, exceeding Python's default recursion limit (sys.getrecursionlimit() = 1000). The emphasis() and strong() methods (lines 80-84) wrap recursively rendered child content in and tags, perpetuating the recursion. This vulnerability affects all mistune APIs including markdown() and html().
Core vulnerable code path:
# src/mistune/renderers/html.py:40-57
def render_token(self, token: Dict[str, Any], state: BlockState) -> str:
func = self._get_method(token["type"])
attrs = token.get("attrs")
if "raw" in token:
text = token["raw"]
elif "children" in token:
text = self.render_tokens(token["children"], state)
else:
if attrs:
return func(**attrs)
else:
return func()
if attrs:
return func(text, **attrs)
else:
return func(text)
The recursive call to render_tokens() on line 48 processes nested child tokens. With 500 levels of nested emphasis/strong tokens, this recursion exceeds Python's default recursion limit of 1000, causing a RecursionError.
# src/mistune/renderers/html.py:80-84
def emphasis(self, text: str) -> str:
return "<em>" + text + "</em>"
def strong(self, text: str) -> str:
return "<strong>" + text + "</strong>"
The emphasis() and strong() methods wrap their text content (which is itself the recursively-rendered output of nested child tokens) in HTML tags, creating the chain of recursion: each call to strong() includes rendered children that themselves call strong(), etc.
POC
from mistune import html
payload = '*' * 1000
try:
result = html(payload)
print('No crash - recursion handled')
except RecursionError as e:
print(f'[VULN] RecursionError: {e} - Process would crash!')
except Exception as e:
print(f'Error: {type(e).__name__}: {e}')
Impact
CVSS 3.1: 7.5 (High), AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H. An attacker can crash any server process using mistune with approximately 2KB of Markdown input. In web applications, this can be triggered through user-generated content such as forum posts or comments, causing denial of service for all concurrent users sharing the same Python process. Both mistune.markdown() and mistune.html() are affected.
CVE-2026-76098 has a CVSS score of 7.5 (High). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (3.3.3); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
- Add a maximum nesting depth limit in the emphasis parsing stage, similar to BlockParser's max_nested_level mechanism (currently at DEFAULT_MAX_NESTED_LEVEL = 20). When the limit is exceeded, treat excess emphasis markers as literal text. 2. Alternatively, refactor HTMLRenderer to use an explicit stack-based iterative approach instead of recursive calls for rendering nested tokens. 3. As a defense-in-depth measure, document the recursion risk and recommend that applications deploying mistune set a higher recursion limit or implement request-level timeouts.
Frequently Asked Questions
- What is CVE-2026-76098? CVE-2026-76098 is a high-severity security vulnerability in mistune (pip), affecting versions >= 3.3.0, < 3.3.3. It is fixed in 3.3.3.
- How severe is CVE-2026-76098? CVE-2026-76098 has a CVSS score of 7.5 (High). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which versions of mistune are affected by CVE-2026-76098? mistune (pip) versions >= 3.3.0, < 3.3.3 is affected.
- Is there a fix for CVE-2026-76098? Yes. CVE-2026-76098 is fixed in 3.3.3. Upgrade to this version or later.
- Is CVE-2026-76098 exploitable, and should I be worried? Whether CVE-2026-76098 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-76098 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-76098? Upgrade
mistuneto 3.3.3 or later.