# knowledge-base
6 articlestagged with “knowledge-base”
Advanced Cloud AI Security Assessment
15-question advanced assessment covering cloud AI attack surfaces across AWS, Azure, and GCP: guardrail bypass, knowledge base exploitation, managed identity abuse, model customization risks, and multi-cloud attack paths.
Bedrock Attack Surface
Comprehensive red team methodology for Amazon Bedrock: model invocation API abuse, guardrails bypass techniques, custom model endpoint exploitation, IAM misconfigurations, knowledge base poisoning, and Bedrock Agents exploitation.
Simulation: Enterprise RAG Security Assessment
Full engagement simulation assessing an enterprise RAG-powered knowledge base for poisoning, exfiltration, and injection vulnerabilities.
Knowledge Base Poisoning (Rag Data Attacks)
Advanced corpus poisoning strategies for RAG systems, including black-box and white-box approaches, scaling dynamics, and the PoisonedRAG finding that 5 texts in millions achieve 90% attack success.
AWS Bedrock Red Team Walkthrough
Complete guide to red teaming AWS Bedrock deployments: testing guardrails bypass techniques, knowledge base data exfiltration, agent prompt injection, model customization abuse, and CloudTrail evasion.
AWS Bedrock Red Team Walkthrough (Platform Walkthrough)
End-to-end walkthrough for red teaming AI systems on AWS Bedrock: setting up access, invoking models via the Converse API, testing Bedrock Guardrails, exploiting knowledge bases, and analyzing CloudTrail logs.