AWS PRIVILEGE ESCALATION: IAM RISKS, SERVICE-BASED ATTACKS, AND NEW AI-DRIVEN BEDROCK/AGENTCORE VECTORS

Summary from the TLDR InfoSec newsletter

AWS privilege escalation has evolved from classic IAM misconfigurations to service-based execution paths and now AI-driven orchestration via Bedrock Agents and AgentCore Code Interpreters. New attack vectors include creating code interpreters that execute Python under privileged agent roles and hijacking Lambda functions backing Bedrock agent tools, both of which enable multi-stage privilege chains to sensitive resources. Testing across 16 scenarios showed that all were exploitable in unrestricted environments but preventable with properly configured SCP-based guardrails, highlighting that native AWS controls often cannot granularly constrain high-risk actions that lack ARNs or condition keys.