Title: AWS re:Inforce 2024 - Cloud upgrade: Modern TLS encryption for all AWS service connections (DAP304)

Insights:

• Session Overview: The session focused on upgrading AWS service connections using modern Transport Layer Security (TLS) protocols, specifically requiring a minimum of TLS 1.2 and enabling TLS 1.3.
• Key Participants: Janelle Hopper (Principal Technical Program Manager, AWS Security), Stephen Colison (Senior Software Development Engineer, EC2 Load Balancing Team), and Mark Ryland (Director, AWS Security).
• Importance of TLS: TLS is crucial for securing connections to AWS service endpoints, impacting all API calls, including those to S3, Kinesis, and EC2.
• Deprecation of TLS 1.0 and 1.1: AWS has removed support for TLS 1.0 and 1.1 due to known vulnerabilities, despite having mitigations in place. This change simplifies audits and aligns with compliance standards.
• Global Rollout: The upgrade to TLS 1.2 or higher was a nearly two-year effort, affecting over 9,000 service public API endpoints and 200 global services. AWS detected 185,000 accounts using outdated TLS versions and notified them to update.
• Performance Improvements with TLS 1.3: TLS 1.3 reduces the number of network round trips required for a handshake, improving connection performance, especially over long network links.
• Compatibility Challenges: AWS faced challenges due to the heterogeneity of clients connecting to their endpoints. They used a phased rollout and monitoring to ensure compatibility and minimize disruptions.
• Monitoring and Rollback: AWS implemented advanced monitoring techniques to detect issues during the rollout and used rollback and sidelining strategies to address compatibility problems without affecting all clients.
• Future of TLS and Post-Quantum Cryptography: AWS is preparing for the future with post-quantum cryptography and the adoption of HTTP/3 and QUIC to further enhance security and performance.

Quotes:

• "We raise the security bar, reduce complexity, and simplify your audits."
• "We rolled this out gradually on a server-by-server basis, doing this almost nearly two years."
• "TLS 1.3 improves this by collapsing those two operations into one by having the client choose a key exchange algorithm it thinks the server will support ahead of time."
• "Our TLS 1.2 implementations and configurations are still hardened against known attacks."
• "We meticulously locked down this data with various security controls to ensure we maintain data privacy."
• "One exciting note I think is per our recent testing, it appears that we are the first major cloud provider to successfully remove these outdated TLS protocols."
• "We found that this is an issue because if there's an issue for a small subset of clients, their failure rate may not meaningfully change the overall success rate into our multi-tenant endpoints."
• "Our approach, whether we were rolling back or sidelining, was to reach out to customers and we have our corpus of client-side risks that we used to give customers actions to update their clients."
• "By adding TLS 1.3 support to our endpoints, it's really raising the security ceiling."
• "The expectation and hope is that you can solve all kinds of new challenges that are, for example, in especially things like modeling nature, right? Physics and chemistry."
• "If we don't start working on this problem now, by the time the risk becomes real, it'll be too late because of the difficulty of upgrading all our systems."
• "We've learned a lot of lessons about things like sidelining and how to do this at scale and so we can apply those same learnings in this case."