I discovered that AWS Bedrock AgentCore Code Interpreter's "Sandbox" mode leaks DNS queries, enabling a full DNS-tunneling C2 channel, reverse shell, and data exfiltration from a supposedly network-isolated sandbox. Disclosed via HackerOne.
Research and proof-of-concept showing how AWS Bedrock AgentCore Code Interpreter's "Sandbox" network mode leaks DNS queries, enabling a full DNS-tunneling C2 channel, reverse shell, and S3/DynamoDB data exfiltration out of a supposedly network-isolated sandbox.
I worked with Barak Schoster to build Cloudsplaining's IAM Scanning logic into Checkov so you can enforce least privilege in Terraform code.
A cheatsheet of SQL queries to use in CloudQuery to identify public network endpoints in AWS.
Short tutorial to set up recurring Prowler scans in AWS
Worried about AWS IAM permissions that allow you to access data (like S3 objects, CodeCommit code, others) or return credentials in their response? I've updated my list of these potentially dangerous API calls and you can read about them here.
This tutorial covers building secure AWS AMIs with Packer. We go over building the prerequisite infrastructure to create EC2 AMIs from scratch, how to use the example hardening scripts for CentOS, and how to validate the testing with Amazon Inspector.
Introduction to Policy Sentry, an easy way to create least privilege policies by copying/pasting ARNs into a YAML file.