Improving Side-Channel Protections for Intel TDX

Описание: "Single-stepping" is a term most often used in the context of debugging, when a software developer tells the debugger to execute a single instruction. In the context of confidential computing, single-stepping can also refer to a privileged adversary (e.g., a malicious OS/VMM) using a timer interrupt or fault to execute a single instruction inside a trusted execution environment (TEE). Repeatedly single-stepping the TEE can allow the adversary to analyze timing and/or power side channels potentially exposed by each instruction executed in the TEE.

Early versions of Intel Trust Domain Extensions (Intel TDX) firmware had a built-in mitigation to address malicious single-stepping. Later, external researchers and internal researchers separately discovered two different methods to bypass this mitigation. Consequently, Intel has updated the single-stepping mitigation in newer versions of Intel TDX firmware. This talk delivers a deep dive on the ineffective single-step mitigation, the two different methods that break the mitigation, and the enhanced mitigation that prevents these malicious single-stepping techniques. Throughout the deep dive we interweave 4 demos that depict a real-world cat-and-mouse game between a malicious VMM and a TDX trust domain.

By:
Scott Constable | Defensive Security Researcher, Intel Labs
Nagaraju Kodalapura | Principal Engineer, Offensive Security Research, Intel Corporation
Baruch Chaikin | Principal Engineer, CPU Architecture, Intel Corporation

Full Abstract and Presentation Materials:
https://www.blackhat.com/eu-24/briefi...