Talk by Adria Gascon (Google)
summary
Presents SNPeek, a practical evaluation of AMD SEV SNP showing nested page tables leak 4 KB granularity memory-access patterns, ciphertext visibility reveals byte-level changes and performance counters expose control-flow. Builds a kernel patch (parfait repo) to collect leakage traces and trains classifiers (LSTM) to reconstruct data from PIR, ORAM and a Stable DP histogram; finds ORAM fails unless aligned to leakage model but DP-style noise partially restores privacy. Proposes ORAM variants, per-app mitigations and defense-in-depth.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:34Confidential virtual machines
- 8:44Threat model and hardware
- 13:05Side-channel attacks
- 20:12Applications and workloads
- 23:20SNPeek and research
- 29:42Leakage analysis and attacks
- 31:09Mitigations and future work
- 33:02Defense in depth and conclusion
related talks
52mSecuring The Unseen: Vulnerability Research In Confidential Computing
25mThe Memory of a Meltdown
25mTelling Your Secrets without Page Faults...
46mRethinking Architectures and Abstraction for a World Where Security Improvements Matter More than Performance Gains
15mI Know What You Said: Unveiling Hardware Cache Side-Channels in Local Large
01h 01mTutorial: Side-Channel Attacks and Transient Execution Vulnerab...
24mThe Memory of a Meltdown
01h 28m“Spectre: Secrets, Side-Channels, Sandboxes, and Security”
