February 10, 2026
Google Cloud Security and Intel conducted a five-month joint security assessment of Intel Trust Domain Extensions (TDX) 1.5, focusing specifically on advanced new capabilities for confidential virtual machines: Live Migration (relocating encrypted confidential VMs between physical host servers without exposing cryptographic keys to the host hypervisor) and TD Partitioning (enabling nested virtualization within Trust Domains).
Key Highlights & Findings
- Vulnerabilities Uncovered: The review identified five security vulnerabilities and approximately 35 architectural edge cases, specification gaps, and hardening opportunities. The most severe flaw could have enabled a malicious Virtual Machine Monitor (VMM / hypervisor) to completely undermine the security guarantees of a Trust Domain.
- Confidential Memory & Speculative Leakage: Additional issues included potential confidential memory leakage from the Intel TDX Module and speculative execution side-channel vectors in guest communication handlers.
- Hands-on Silicon & Novel Tooling: Unlike earlier design-only reviews, the team executed attacks directly on early TDX-enabled hardware nodes, developing custom proof-of-concept testing harnesses (the TDXplore Toolkit).
- AI-Assisted Security Analysis: The assessment leveraged Gemini for automated code analysis and NotebookLM to cross-reference and verify intricate multi-thousand-page architectural specifications.
- Remediation: Intel fully addressed and mitigated all identified vulnerabilities in updated TDX 1.5 microcode and firmware releases prior to production cloud rollout.
📖 Read the full report on Google Bug Hunters:
Strengthening the Foundation: A Joint Security Review of Intel TDX 1.5
📄 Full Technical Paper: Security Assessment of Intel TDX with Support for Live Migration (PDF)