Ep053: Building Generative AI Powered Security Assistants with Amazon Bedrock

Lacework’s VP and Head of Engineering Arash Nikkar, and Product and Security Software Engineer Teddy Reed, discuss their work with generative AI and Amazon Bedrock, focusing on threat detection, chat interface improvements, and the rapid development of AI-driven solutions.

Topics Include:

• Introducing Lacework with Arash Nikkar
• Working with AWS and Bedrock
• Focusing on threat detection
• Lacework ingests over a trillion events each day
• Composite alerts for detecting anomalies
• Developing and improving the chat interface
• Teddy Reed talking chat interface improvements
• Reviewing the architecture
• Fast engagement with generative AI / AWS Bedrock
• The rapid delivery with Bedrock – took 20% of expected time
• Best practice – build in time to review and verify responses
• Measuring customer’s engagement and return frequency and feedback
• Understanding customer sentiment and topic analysis
• Achieving 60% completed results and feedback
• Transparency of source, response
• Enabling assistant to have more access to data from LLM
• Using AWS’ model evaluator, raising the accuracy scores
• Moving everything over to Bedrock for multiple reasons
• Familiarity of AWS tools helpful for all development and security teams
• Always challenging assumptions – is chat interface the right interface?
• Q&A 1: Using vector database for analysis, using results as system prompt
• Q&A 2: Approaching cost-optimization and balancing ROI
• Q&A 3: Using as automated remediation for findings
• Q&A 4: Data source Bedrock in leveraging

Session wrap up

Participants:

• Arash Nikkar – VP, Head of Engineering – Lacework
• Teddy Reed – Product and Security Software Engineer - Lacework