Stanza: RCE via pickle deserialization in the lemmatizer
Stanza 1.14.0 before 1.15.0 can run arbitrary code as the loading user when a caller-supplied lemmatizer checkpoint is loaded. Fixed in 1.15.0.
Written by anthropic/claude-haiku-5.5 from the sources below. Verify against them.
- Exploited
- No known exploitation
- Fix
- Fixed in 1.15.0
- Who is exposed
- Specific setups only
Do first: Check whether your setup matches what the bug needs. Stanza version 1.14.0 (lemmatizer stores pos_dict as gzip-compressed pickle) A caller-supplied lemmatizer checkpoint is loaded, such as through a pipeline built with a lemma model path
- Exploitation0/40No signal
- Exposure5/25Specific setups only
- Reach12/20stanza: 255K downloads a week
- Severity8.8/10CVSS 7.8, total technical impact per CISA
- No patch0/5No signal
What would change the score
- Would score about 76 if added to CISA KEV
- Would score about 71 if CISA's SSVC rates exploitation active
- Would score about 36 if CISA's SSVC rates it automatable
CVE-2026-115878 is not in CISA's Known Exploited Vulnerabilities (KEV) catalog. No public exploit code is tracked. It is rated CVSS 7.8 (high) and affects stanza; fixed in 1.15.0.
- Last modified
- Where the record stands: published, rejected by its publisher, or superseded by another id.What do these mean? →
- Published
- Whether CISA lists it in its Known Exploited Vulnerabilities catalog: proof that attackers are using it.What do these mean? →
- Not listed in CISA KEV
- A 0-100 score for how much a CVE matters across the industry: exploitation, exposure, how widely the software is used, severity and whether a fix exists.What do these mean? →
- 26/100 · provisional
- How many real-world setups can hit the bug, from the vendor's advisory and CISA data. It does not know your setup.What do these mean? →
- Specific setups only
- Fix
- Fixed in 1.15.0
- Whether working attack code is public (Exploit-DB, Metasploit, Nuclei or GitHub). Public code makes attacks easier.What do these mean? →
- None on Exploit-DB, Metasploit or GitHub
- Exploitation: none
- Automatable: no
- Technical impact: total
Summary
Stanza, a Stanford NLP Python library for processing human language text, is affected when loading lemmatizer model checkpoints. An attacker who supplies a malicious lemmatizer model file can run arbitrary code on the system during pipeline setup, before any text is processed.
Read the original description
Stanza is a Stanford NLP Python library for tokenization, sentence segmentation, NER, and parsing of many human languages. From 1.14.0 until 1.15.0, loading an attacker-supplied lemmatizer checkpoint can execute arbitrary code because stanza/models/lemma/trainer.py stores pos_dict as gzip-compressed pickle bytes and Trainer.load passes the decompressed stream to unrestricted pickle.loads. The outer torch.load call uses weights_only=True, but the nested bytes are treated as a primitive and are not inspected by that allowlist. The path is reachable during pipeline construction through a caller-influenced lemma_model_path, before inference begins. This issue is fixed in version 1.15.0.
Impact
- Widely used
stanza has 255k weekly downloads.Source: pypistats.org
103 packages depend on stanza.Source: deps.dev
What to do
Check whether your setup matches what the bug needs.
Stanza version 1.14.0 (lemmatizer stores pos_dict as gzip-compressed pickle) A caller-supplied lemmatizer checkpoint is loaded, such as through a pipeline built with a lemma model pathUpgrade to a fixed version: 1.15.0.
Timeline
- CVE published
- First seen by cve.now
- CISA SSVC assessment
Exploitation: none · Automatable: no · Technical impact: total
- Record last updated
Affected versions
- stanfordnlp stanza
1.14.0
- stanzaPyPI
From 1.14.0 · fixed in 1.15.0
Evidence 4
- Advisory
- Repository
- Repository
- Repository
Corrections
None so far.
Something wrong on this page? Report an error. Every fix is listed here.
Questions about CVE-2026-115878
Is CVE-2026-115878 being exploited?
It is not in CISA's Known Exploited Vulnerabilities catalog, and CISA's SSVC assessment does not mark exploitation as active. That does not mean it has not been attacked: vendors and researchers can report attacks before CISA lists them. No public exploit code is tracked.Source: cve.now feed data
What does this mean?
The lemmatizer stores its pos_dict as a gzip-compressed pickle, and loading a lemmatizer checkpoint (.pt file) with attacker-supplied contents runs arbitrary code as the user who loaded it.Source: github.com, github.com
Which versions are affected by CVE-2026-115878?
stanza: version 1.14.0. stanza (PyPI): from 1.14.0 before 1.15.0.Source: cve.now feed data
Am I affected, and how do I check?
Stanza versions from 1.14.0 up to but not including 1.15.0, in pipelines that load a caller-supplied lemmatizer checkpoint, such as one built with a lemma model path. To check, a reader can print the installed version with pip show stanza and review whether the lemma model path points to a custom checkpoint.Source: github.com, github.com
How do I fix CVE-2026-115878?
Upgrade to Stanza 1.15.0, which removes the code path that loads old models.Source: github.com, github.com