Securing Tomorrow’s Hardware: Dual-State Models in the AI Era
summary
Claims that a dual-state model confines secrets by enforcing a hardened secure zone separate from a non‑secure runtime, backed by concrete mechanisms: address space partitioning, secure boot chains, periodic runtime integrity checks, and AI-driven anomaly detection. Examples cite Spectre/Meltdown class leaks, hardware enclaves (Arm TrustZone and Intel SG), and supply‑chain controls like secure device identifiers, tamper‑evident packaging, and attestation. Emphasis falls on implemented tools and measurable defenses rather than vague promises.
AI-written summary · about
chapters (9)
- 0:00Introduction and Speaker Background
- 0:15The Importance of Hardware Security in AI
- 0:54AI's Unique Vulnerabilities
- 3:22Address Space Partitioning
- 5:28Secure Boot and Runtime Integrity
- 7:59AI-Driven Anomaly Detection
- 13:37Dual State Security Models
- 15:23Supply Chain Security
- 16:40Future Challenges and Conclusion
related talks
46mRethinking Architectures and Abstraction for a World Where Security Improvements Matter More than Performance Gains
32mBeyond Software: Hardware-Based Security
38mOne Ring -3 To Secure Them All: Computing with Hardware Enclaves
01h 13mHardware-Software Contracts for High Assurance with Applications to Side-Channel Security
30mKeynote Panel: The Future of...
36mSecurity for AI and Multi-Party Collaboration with Confidential Computing (Oh, and...
44mHardware Security By Design
23mHardware security for critical national infrastructure