# knowledge-base
6 artikelengetagd met “knowledge-base”
Geavanceerd beveiligingsassessment van cloud-AI
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.
Aanvalsoppervlak van Bedrock
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.
Simulatie: beveiligingsbeoordeling van enterprise-RAG
Full engagement simulation assessing an enterprise RAG-powered knowledge base for poisoning, exfiltration, and injection vulnerabilities.
Poisoning van de kennisbank (RAG-data-aanvallen)
Geavanceerde strategieën voor corpus-poisoning van RAG-systemen, inclusief black-box- en white-boxbenaderingen, schaaldynamiek, en de PoisonedRAG-bevinding dat 5 teksten op miljoenen een aanvalssucces van 90% behalen.
Walkthrough: AWS Bedrock red team
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.
Walkthrough: AWS Bedrock red team (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.