I’ve been thinking about software and hardware backdoors recently.
The most important protection is first to have a good control over any connection, wired, wireless, electromagnetic etc. so that the compromised system cannot attempt any connection on its own.
But then, the backdoor could alter the safe/allowed connections to behave in malicious ways being undetected. For example, in context of anonymity the backdoor could somehow fingerprint or inject identifiers into connections.
Another major topic is encryption and random entropy, I guess we don’t rely on a single source, any improvements possible here?
I know thic topic is mostly theoretical but does anyone know how to protect against known or possible data interferences? Is it possible to have a MAT-like automatic identifier stripper on Whonix-Gateway, on host, on router, etc?
Other than protecting against backdoors, could we have such tools based on some updated database to delete known identifiers from Workstation connections? For example, for protecting anonymity of Whonix users against poorly/not configured (popular) applications (hidden services would be a great example) installed on Workstation. Or in other words protect any idiot’s or newbie’s anonymity, or simply have an additional protection just in case
Entropy:
Whonix is using standard sources + haveged. See:
Contributions/improvements welcome!
Hardware backdoors:
If they exist and are used, we’re hosed. Simple as that. This is probably outside the scope of Whonix. I guess many funds (https://www.opentechfund.org/submit/alternative-sources-support) and/or other investors could get excited if one seriously planing to create hardware, that can be vetted to be backdoor free.
Software backdoors in Whonix itself:
I think Whonix isn’t doing so bad here. Check out:
There is many BIOS threat attack vectors that exist:
Intel Management Engine/AMT/vPro
CompuTrace/Absolute/LoJack
PXE Boot/Network Boot
DXE Modules/Drivers
NVRAM Persistence
Would be interested in a Wiki section about mitigation’s and am familiar with a few would be glad to help. The last one is a bit tricky since simply removing CMOS on modern UEFI doesn’t clear all settings like bios password etc and reset defaults.
Without going into a super deep and technical discussion, just because RISC-V is an open ISA does not mean DeepComputing and/or other vendors will implement a fully open stack. For example, Framework will be releasing another update of the Framework Laptop 13 with an alternative RISC-V mainboard:
Even if RISC-V achieves feature and/or performance parity with x86, vendors could still simply implement proprietary firmware and hardware into the silicon itself. Furthermore, even if the entire stack is completely open, that does not mean it will remain open for every future revision, which is exactly the siloed trap Raptor Computing Systems is still caught in from POWER9 to Power10, along with the rest of the FSF movement.
hope you don’t get the "Amazon special"™