CVE-2026-128139: NLTK: XMLCorpusView comment and CDATA parsing permits quadratic CPU denial of service
High CVSS 7.5
- OSV id
- GHSA-j8g8-j4j7-8j54
- Also known as
- CVE-2026-128139
- Published
- Last modified
Affected packages
| Package | Affected versions | Fixed in |
|---|---|---|
| nltk PyPI |
>= 3.6, < 3.10.4 | 3.10.4 |
How to fix it
Upgrade nltk to 3.10.4 or later.
# in requirements.txt: change the line to nltk==3.10.4
pip install -r requirements.txt
- If it is an indirect dependency, upgrade the package that depends on it, or add a minimum version constraint for it (e.g. in constraints.txt, or as a dependency in pyproject.toml).
- Check the changelog for breaking changes (especially across major versions), update the lockfile with the package manager, and run your tests.
Not sure whether your project uses an affected version? Check your lockfile below for free.
Details
affected: PyPI
nltk>= 3.6—nltk.corpus.reader.xmldocs.XMLCorpusViewstreaming read (and thenps_chat/framenetreaders built on it) dynamically reproduced:3.6(6/6 checks),3.9.2(6/6),3.10.3(6/6) anddevelop6818767b9(6/6)
Source references (nltk/corpus/reader/xmldocs.py:NNN) are to commit 6818767b99c24659107eb2be9bccd32a3d4bf045 (develop, v3.10.3-15-g6818767b9). The released v3.10.3 wheel's xmldocs.py is byte-identical to that commit, so the same line numbers apply to the released wheel.
Summary
An attacker who can make a victim read a well-formed XML corpus file whose body contains one large comment or CDATA section can pin a CPU core for a time quadratic in the file size, because XMLCorpusView._read_xml_fragment (nltk/corpus/reader/xmldocs.py) re-runs a linear whole-buffer validation after every 1 KiB block while the comment/CDATA terminator is still ahead. XMLCorpusView is NLTK's documented streaming API for XML corpora and is instantiated directly by the NPSChatCorpusReader and FramenetCorpusReader readers; the module header (xmldocs.py:16-20) explicitly handles untrusted corpus XML. The in_tag guard added to this function for GHSA-vp2x covers only a single oversized <...> tag and leaves the CDATA/comment path quadratic in the current release.
Details
Root cause: _read_xml_fragment (xmldocs.py:273-338) accumulates the file into one fragment string in _BLOCK_SIZE (1024) character chunks (xmldocs.py:123, 298-299). After every block that is not yet a well-formed prefix, the fragment is re-validated against _VALID_XML_RE. Two whole-buffer operations dominate:
# xmldocs.py:308
if not in_tag and self._VALID_XML_RE.match(fragment): # whole-buffer match (O(len))
return fragment
# xmldocs.py:312
if re.search("[<>]", fragment).group(0) == ">": # O(1) for this input class:
... # the document starts with '<'
# xmldocs.py:329
if self._VALID_XML_RE.match(fragment[:last_open_bracket]): # whole-buffer prefix match (O(len))
...
For this input class the re.search probe at xmldocs.py:312 is O(1): the fragment always starts with < (e.g. <root>…), so the first </> character is at offset 0 and it never contributes cost. The quadratic cost comes entirely from the two whole-buffer regex matches at xmldocs.py:308 and xmldocs.py:329 (the latter scans the buffer up to the most recent <, which for this input sits near the end of the buffer).
_VALID_XML_RE (xmldocs.py:238-249) is a linear regex (its tempered-dot comment/CDATA bodies (?:(?!-->).)* / (?:(?!\]\]>).)* scan left-to-right in O(length) — confirmed by the embedded script's single-match check), but for a fragment whose terminator is far ahead the match itself fails after scanning the whole buffer. With an unclosed-across-blocks CDATA or comment, every one of the n/1024 blocks pays O(current buffer length) on lines 308 and 329, for a total of ~O(n²) character operations.
The in_tag guard added for GHSA-vp2x (xmldocs.py:283-305) tracks only bare </> characters: it is True inside an unclosed <...> tag and, in that state, skips only the line-308 match. It never represents "the fragment sits inside an unclosed CDATA or comment" — XML allows arbitrary </> inside those constructs — so both the line-308 match and the line-329 backtrack match still fire on every block for a CDATA/comment fragment. The comment block at xmldocs.py:283-289 even documents the intended linearity contract ("re-running it on every 1 KiB block otherwise makes a single oversized tag O(n^2) (CWE-407)"); the guard implements it for the tag branch only.
The single-match cost is linear, not super-linear: the embedded script measures _VALID_XML_RE.match() on 512 KiB → 1 MiB → 2 MiB unclosed-CDATA buffers and gets ≈ 0.074 s → 0.142 s → 0.280 s (≈ 1:1.9:3.8, i.e. linear), asserting the 4× ratio stays well below the quadratic bound. The quadratic total therefore comes from re-running those linear matches per block — the exact CWE-407 pattern the in-tag guard was introduced to eliminate.
Version nuance (all dynamically verified):
3.10.3/develop(guard present): the pinned regex validates CDATA correctly, so both the comment and the CDATA attack documents read successfully and return; both scale super-linearly (t(4n)/t(n) ≈ 13-16; linear would be ≈ 4). The oversized single-tag control reads linearly (ratio ≈ 4.0), proving the quadratic cost is specific to the unguarded CDATA/comment path.3.6/3.9.2(pre-guard): these releases use a lazy_VALID_XML_REwhose CDATA alternative<![CDATA[.*?]]>never matches (the[CDATA[.*?]substring is parsed as a character class), so a CDATA document is rejected withValueError: Unexpected end of file: tag not closedafter the quadratic scan — a separate functional defect. The comment variant i
References
- GHSA-j8g8-j4j7-8j54 on OSV.dev
- Web: https://github.com/nltk/nltk/security/advisories/GHSA-j8g8-j4j7-8j54
- Advisory: https://nvd.nist.gov/vuln/detail/CVE-2026-128139
- Web: https://github.com/nltk/nltk/commit/9cacbb95d97eaac7debb30d6e7decbe74d2c635f
- Package: https://github.com/nltk/nltk
- Web: https://github.com/nltk/nltk/releases/tag/v3.10.4
Am I affected?
Upload or paste your lockfile. We check every package version in it against OSV.dev, this advisory included, and email you the vulnerable ones with the version that fixes them. Free, and no account needed.
About this data
Source: GitHub Advisory Database via OSV.dev (CC-BY 4.0). Severity comes from the advisory's own rating when it has one, else from its CVSS score.