Skip to content
Submit incident
AII-1002-44867
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.

MindsDB Weaviate integration allows arbitrary code execution via eval injection

A vulnerability in MindsDB's Weaviate integration allows authenticated attackers to execute arbitrary code by injecting Python code through embedding vector filter parameters. The flaw affects MindsDB versions v23.10.3.0 through v24.7.4.1.

12 September 2024 Β· MindsDB Β· Security and cyberattacks
12 September 2024
AII-1002-44867View source β†—
Quick facts
Category
Security and cyberattacks
Source
hiddenlayer.com
Date
12 September 2024
Drafted by
Raidu Incident Detection

What happened

A vulnerability in MindsDB's Weaviate integration allows arbitrary code execution through improper neutralization of directives in dynamically evaluated code. An authenticated attacker can exploit the flaw by injecting Python code into the embedding vector filter parameter of a SELECT WHERE clause.

The vulnerability exists in the mindsdb/integrations/handlers/weaviate_handler/weaviate_handler.py file where the eval function is used to parse the embeddings column argument without proper sanitization, allowing arbitrary Python code execution.

Index analysis

The vulnerability affects MindsDB versions v23.10.3.0 through v24.7.4.1 and has a CVSS score of 8.8, indicating high severity. Exploitation requires authentication to a MindsDB instance with the Weaviate integration installed and the presence of an embedding vector filter.

Classification

Category
Security and cyberattacks
Industry
IT infrastructure and hosting
Harm status
Potential harm
Intent
Unintentional
Timing
Unknown
AI systems
MindsDB, Weaviate integration

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

hiddenlayer.com
https://hiddenlayer.com/sai-security-advisory/2024-09-mindsdb/