Secure Amazon Bedrock AgentCore Runtime with AWS WAF
▶ The 2-minute explainer
Summary
This post details two architectural patterns for securing Amazon Bedrock AgentCore Runtime using AWS WAF, an internet-facing Application Load Balancer (ALB), and a VPC Interface Endpoint. It covers both a Lambda-proxied approach for request transformation and a direct ALB-to-VPC Endpoint method, along with steps to enforce WAF-only traffic via resource policies and tested authentication methods.
Why it matters
For professionals deploying AI applications on AWS, understanding these security patterns is crucial for protecting sensitive data and ensuring compliance by properly securing access to AI services like Amazon Bedrock.
How to implement this in your domain
- 1Evaluate your security requirements for Amazon Bedrock AgentCore Runtime deployments.
- 2Choose between the Lambda-proxied or direct ALB-to-VPC Endpoint pattern based on your need for request transformation.
- 3Configure an internet-facing Application Load Balancer (ALB) and integrate it with AWS WAF for traffic filtering.
- 4Set up a VPC Interface Endpoint to privately connect your ALB to the AgentCore Runtime.
- 5Implement a resource policy to restrict direct access to AgentCore Runtime, forcing all traffic through AWS WAF.
Who benefits
Key takeaways
- AWS WAF is essential for securing Amazon Bedrock AgentCore Runtime.
- Two primary architectural patterns exist for secure deployment.
- Lambda proxy offers request transformation control; direct method is simpler.
- Resource policies can enforce WAF-only traffic.
Original post by Puneeth Komaragiri
"This post shows you two architecture patterns that address this problem. Both use an internet-facing ALB with AWS WAF and route traffic through a VPC Interface Endpoint to AgentCore Runtime. Pattern 1 places an AWS Lambda proxy between the ALB and the VPC Endpoint, giving you ful…"
View on XOriginally posted by Puneeth Komaragiri on X · view source
Want to go deeper?
Turn these trends into skills with Learnijoy's hands-on AI & tech courses.
Explore coursesMore in AI Engineering & DevTools
LingBot-World 2.0 Infinity: New Interactive AI World Model Released
LingBot-World 2.0 (Infinity) has been released on Hugging Face, offering an interactive world model capable of hour-long generation without quality degradation, featuring rich actions and an agentic director for real-time evolution.
LingBot-Video: New MoE Video Foundation Model Released
LingBot-Video, a new Mixture-of-Experts (MoE) based video foundation model, has been released on Hugging Face, designed for embodied intelligence with 30 billion parameters and augmented with extensive embodied data.
RynnWorld-4D: New 4D Embodied World Models for Robotics
RynnWorld-4D introduces new 4D embodied world models specifically designed to enhance robotic manipulation capabilities, with an accompanying research paper available.