Elektrine lite

← Feed

@david_chisnall@infosec.exchange

2026-09-26 08:14 UTC

@vmcall@infosec.exchange @notbobbytables@infosec.exchange When we did the ThunderClap attacks (ring-0 arbitrary-code execution by plugging something into the Thunderbolt port), the response from the Windows team was that it was out of scope for their threat model. The response from the OS X team was to tighten IOMMU permissions and to rearrange the mbuf structure so that the part exposed for DMA doesn’t contain pointers. The worst that we could do after that was corrupt data being loaded to the buffer cache from the disk if we forged the PCIe identifier and timed it exactly right (and somehow knew the physical address that was being written to and when), and that was closed by subsequent iterations of the hardware.

Replies (0)

No replies.