In order for me to trust AMD's implementation, they first need to can that ridiculous Platform "Security" Processor. It is as useless and dangerous as Intel Management Engine, running unknown code.
A more plausible attack would be an application using malloc() for a large segment of memory, and transmitting the "uninitialised" content, which could contain private keys, sensitive documents, etc. from applications that either don't zero the memory after finishing, or programs which have crashed and the memory is now freely available to other processes. It would be nice in those cases to have different keys for different pages, so that when a process is terminated, the kernel can (instruct the CPU to) overwrite the key with a new random number. On Sat, 11 Mar 2017 20:18:37 +0000 (UTC) Christian Weisgerber <na...@mips.inka.de> wrote: > AMD thinks so. Last year they announced support for memory encryption > in future CPUs. The top two Google hits: > > http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/12/AMD_Memory_Encryption_Whitepaper_v7-Public.pdf > https://events.linuxfoundation.org/sites/events/files/slides/AMD%20x86%20Memory%20Encryption%20Technology%20LSS%20Slides.pdf
