Live › vulnerability › CVE-2026-120334
deepset FARM up to version 0.8.0 has a path traversal flaw in EmbeddingTokenizer.from_pretrained, where a manipulated vocab_filename argument is not restricted. A remote attacker can use it to reach files outside the intended directory, and a public exploit has been published.
Summary written from the sources below. Check the originals before acting.
First seen by InfraStack 2026-11-02 03:31 UTC. All times on this page are UTC.
Applications that use FARM's EmbeddingTokenizer.from_pretrained and pass it a value an outside party can influence, such as a request parameter or uploaded configuration. The sources say the vendor did not respond and the affected versions are no longer supported by the maintainer. Teams that load tokenizers only from fixed, trusted paths are not exposed through this flaw.
No product ranges from the sources yet.
0.8.0 or lower.EmbeddingTokenizer.from_pretrained that pass user-supplied values as vocab_filename.No fix yet, per the sources.
Each row is the first time that source carried this signal. The analyst’s rows are its own findings.
A flaw has been found in deepset-ai farm up to 0.8.0. This impacts the function EmbeddingTokenizer.from_pretrained of the file farm/modeling/tokenization.py of the component Tokenizer Loader. Executing a manipulation of the argument vocab_filename can lead to path traversal. The attack may be performed from remote. The exploit has been published and may be used. The vendor was contacted early about this disclosure but did not respond in any way. This vulnerability only affects products that are no longer supported by the maintainer.
This page as .md · .json · in your agent: get_signal("CVE-2026-120334")
NVD data is provided by NIST and is public domain. Posts quoted belong to their authors.