When using any cryptographic solution, the endpoints used must be secure. Malign software/exploits/bugs risk the entire security of any crypto system and forfeits any guarantees made by the software running on commodity PCs.

Considering that certain TLAs (Three Letter Agencies) and FLAB (Four Letter Abbreviated Bureaus) are on all levels an APT, it seems not that far out there that they might be inside of practically every piece of hardware out there.

I hear people talk about having a dedicated key management machine. But this is only used to generate new keys and revoke existing ones, not to encrypt data?

So why is it that I don't see people having dedicated airgapped crypto machines with validated hardware on their desk? Inconvenience? Price? Am I missing something here? Are devices like this out there that I don't know about? Is the security gained by validated hardware somehow less significant than the current ease with which malign code can be injected in open source software?

  • Validated by whom? If the illuminati are as powerful as you suggest then how do you know which is the red pill?
    – symcbean
    Mar 5, 2017 at 22:17
  • +Xunie You may as well assume that all hardware after 2003 has embedded standalone ROMs installed that act as a security breach separate from anything that BIOS or your OS will detect. Look into Jacob Appelbaum's public talk on NSA tools that were leaked and the methods they employ to "hack" hardware before it is even shipped to retail stores. You can find the talk on youtube.
    – Yokai
    Mar 5, 2017 at 23:04

1 Answer 1


You certainly can have dedicated hardware (an HSM, or hardware security module) that both protects keys, and performs cryptographic operations such as encryption, decryption, and signing in a hardened, special purpose appliance.

The primary reason they're not more widely used is, as you have already suspected, price. They're expensive and complicated, and the risk of compromise from a three or four letter agency (or other threat actors that would compromise a cryptosystem) is, to the average organization, not sufficient high as to justify the cost.

You also have to consider that the threat model they protect against is quite small. Even if you're performing cryptographic operations on dedicated hardware, if the hardware communicating with your HSM (sending plaintext to be encrypted, or receiving decrypted ciphertexts) is compromised, the raw communication is still at risk...You're still only ultimately protecting the keys. This does add a valuable layer of security, but it is still a limited layer of security. This has to factor into the equation of whether the risk is limited to a degree that justifies the price. In many cases (or most cases, perhaps) it does not. The performance you can get by offloading crypto operations to dedicated hardware may justify the price, but in many (or perhaps most) cases, the security that you get will not.

  • But what about more common threats? If I don't trust my windows installation to be secure, then I can't trust my UEFI/BIOS either! (which can easily be flashed from Windows) Which now means I can't trust Linux anymore either because my BIOS might be compromised if I use Windows and don't validate every piece of software in the chain. This doesn't sit right with me. It feels like any security precaution I take is simple a facade.
    – Xunie
    Mar 7, 2017 at 1:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .