Data Poisoning - Bedrock Knowledge Base

How to poison Amazon Bedrock Knowledge Base

Overview

• Amazon Bedrock Knowledge Bases can leverage data from multiple sources e.g., S3, Salesforce, and more

• This data may support critical operations such as customer service agents, internal references, and more

• If attackers can gain access to and/or modify the data, it can impact these workloads

Resync Data Source

• Once data is modified (e.g., an attacker corrupts data in an S3 bucket), the knowledge base will need to resync with the data source

aws bedrock-agent start-ingestion-job --knowledge-base-id <knowledge-base-id> --data-source-id <data-source-id>

• Get knowledge base id from aws bedrock-agent list-knowledge-bases

• Get data source id from aws bedrock-agent list-data-sources --knowledge-base-id <knowledge-base-id>