Link following in Natural Language Toolkit - #VU141416
Published: August 11, 2026
Vulnerability details
The vulnerability allows a local user to disclose sensitive information.
The vulnerability exists due to improper link resolution before file access in IPIPANCorpusReader _get_tag() when processing caller-supplied fileids that reference a symlink within the corpus root. A local user can place a symlinked file in a reachable corpus directory and trigger file reading through channels(), domains(), categories(), or fileids(channels=...) to disclose sensitive information.
Literal ../ traversal in the fileid is still blocked, so the issue is specific to symlink-based access to files outside the corpus root.