erroneo.us
Search all guides⌘K

Digest · Monday

Security digest: October 9, 2023

1 editorial brief and 1 critical advisory.

All digests · RSS · Markdown

Take stock

  • Check exposure: LangChain PALChain: Prompt Injection to RCE (CVE-2023-44467). Teams whose Python apps import langchain_experimental and use PALChain or other chains that execute model-generated code on user-influenced input.
  • PyPI: 1 critical advisory (langchain-experimental). Check with pip-audit (or: pip show <package>).
  • None of it touches your stack? Then there is nothing to do today. That is a result, not a failure to read carefully.

Editorial briefs

Check exposure

· AI agent security · CVE-2023-44467, CVE-2024-27444

LangChain PALChain: Prompt Injection to RCE (CVE-2023-44467)

Only apps that feed untrusted text into PALChain (or similar LLM-to-Python chains) are at risk, but for them it is full code execution, and blocklist fixes kept getting bypassed, so remove the pattern rather than trust the patch.

Must act: Teams whose Python apps import langchain_experimental and use PALChain or other chains that execute model-generated code on user-influenced input.

Can relax: LangChain users who never install langchain-experimental or never use its code-executing chains, and users of other LLM frameworks without eval-style features.

  1. Find usage: grep -rn 'PALChain\|langchain_experimental' your code and run pip show langchain-experimental
  2. Upgrade langchain-experimental to the latest release (0.0.52 at minimum)
  3. Stop passing untrusted or retrieved text into PALChain or any chain that executes LLM-generated code

Read the brief →

Known exploited (CISA KEV) 0

Added to CISA’s catalog on this day: confirmed exploitation in the wild. Due date = CISA’s deadline for US federal agencies.

None.

Critical advisories by ecosystem 1

GitHub-reviewed advisories rated critical, published this day (Pacific time).

PyPI 1

Check: pip-audit (or: pip show <package>)