Skip to main content

Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection')

CVE-2023-29374

High
9.8/10

Summary

In LangChain, the "LLMMathChain" chain allows Prompt Injection attacks that can execute arbitrary code via the Python "exec" method. This issue affects the versions prior to 0.0.142.

  • LOW
  • NETWORK
  • HIGH
  • UNCHANGED
  • NONE
  • NONE
  • HIGH
  • HIGH

CWE-74 - Injection

Listed as the number one web application security risk on the 'OWASP Top Ten', injection attacks are widespread and dangerous, especially in legacy applications. Injection attacks are a class of vulnerabilities in which an attacker injects untrusted data into a web application that gets processed by an interpreter, altering the program's execution. This can result in data loss/theft, loss of data integrity, denial of service, and even compromising the entire system.

References

Advisory Timeline

  • Published