NLTK: Stable FrameNet and NKJP readers parse outside-root XML

Published September 8, 2026 CVE-2026-62385

NLTK apps can expose XML files stored outside their assigned data folder. The risk requires attacker-influenced input reaching specific FrameNet or NKJP features.

Severity
MediumCVSS 3.1 · 5.9
Fix
Fixed in 3.10.0Fix recorded 3 days ago
Affected versions
before 3.10.0
Weakness
CWE-73
Exploit likelihood
0.37% in 30 daysEPSS, higher than 30% of known flaws
Affects
NLTK

How it works

  • An application must expose NLTK's FrameNet or NKJP reading features.
  • It must accept attacker-influenced selections, file identifiers, or lookup information.
  • NLTK can turn those values into filenames pointing outside the assigned folder.
  • The application may then read and return accessible XML files from elsewhere.

What to do

Run python3 -m pip show nltk in the application's Python environment. Version 3.9.4 or earlier is affected. Finding the package alone does not prove exposure. Confirm whether the application uses FrameNet or NKJP readers with attacker-influenced input.

Run this in the application environment you want to check:

python3 -m pip show nltk

Update NLTK through the application's dependency process to version 3.10.0 or later. Then repeat the version check in every Python environment running the application.

Technical details

CVE-2026-62385 affects FramenetCorpusReader and NKJPCorpusReader in NLTK versions older than 3.10.0. Several public methods can build file locations from unsafe selectors, file identifiers, or altered lookup data. Those locations can escape the intended corpus folder and expose accessible XML files. Version 3.10.0 rejects file locations that leave the trusted folder.

Severity breakdown

  • Attack vector Network Requires network access to the vulnerable service
  • Attack complexity High Attack depends on conditions outside the attacker's control
  • Privileges required None Attacker needs no account or login
  • User interaction None No victim action needed
  • Scope Unchanged Impact stays within the same security authority
  • Confidentiality impact High Sensitive data can be exposed with serious impact
  • Integrity impact None No data tampering
  • Availability impact None No availability impact
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N Open in FIRST.org calculator

References