daily cyber × ai intelligence

index

tagged

[confidential-computing]

2 editions · 2 items

September 15, 2026

  • DDrop defeats confidential-computing integrity by silently discarding DDR5 writes. Researchers used a sub-$200 interposer to exploit missing memory freshness in Intel TDX, Scalable SGX, and AMD SEV-SNP, recovering private VM memory and switching a victim into debug mode. The lab attack documented by The Hacker News requires host-software control plus brief physical access; researchers report no use outside a laboratory. The team says it is releasing board designs, firmware, and attack code, with the paper due at ACM CCS in November. · Exploitation & Vulnerability Research

in Scope Questions Recast Anthropic’s “Rogue Agent” Incidents

July 5, 2026

  • Confidential computing's core trust mechanism is reported broken, with no clear fix. New research (surfaced by NCSC-FI) claims remote attestation — the cryptographic proof that data is entering a genuine, unmodified TEE such as Intel TDX — rests on a fundamental architectural flaw, calling into question its use as the security foundation for EU sovereign-cloud ambitions. The Register · Vulnerabilities & Exploits

in Confidential Computing's Root of Trust May Be Unfixable