3 ms·

The price seems reasonable for the kind of effort that compliance requires. Though if you are not bound by this very particular security level, it probably does

The price seems reasonable for the kind of effort that compliance requires. Though if you are not bound by this very particular security level, it probably doesn't make sense to use this instead of strong cryptography alone. If the encryption is strong enough, you don't need the fancy intrusion detection.

andai 6y ago
Say, if encryption isn't done on the device itself, then it wouldn't need to be tamper proof would it?
- solarkraft 6y ago
  Yeah. To my untrained eye it looks like FIPS-140 was designed for devices handling unencrypted data (anything with computation) where physical security matters and this USB stick is just a by-product of a blanket requirement that it's used.
  - toyg 6y ago
    Isn't that a good thing? Anything implemented in software puts some burden on the user's side ("are you using the right OS with the right program and the right options in the right way? Just one misstep and it's all for naught"). Offloading more of that cognitive load and operational risk onto dedicated hardware is surely a good thing (if uneconomical) in practice.
    - Nullabillity 6y ago
      The USB host still has access to both the plaintext and key (once authenticated), so they both need to be fully trusted anyway. This garbage just adds yet another link to the critical chain. Oh, and now you have full plain text access over the (unencrypted and unauthenticated) USB bus. Hope you put as much effort into vetting your USB hubs!
      - toyg 6y ago
        When I was writing that post, I had actually added a second paragraph, that basically said "I hope there were some standard mechanisms for the hardware to refuse mounting without a password or some other verification system". I eventually deleted it because there might well be something of the sort already, in the bowels of TPM specs and similar. There is no reason, imho, why we shouldn't strive to "encrypt all the things", including buses, so that data is guaranteed to be available only to fully-trusted chains. This should obviously be configurable by administrators, with destruction guarantees when options are relaxed. This will not be for everyone (and yes, it will likely always be weak against the $5-wrench attack), but for people who really need it (like deployed troops, whose default opsec practices are typically too lax to be left to humans).

andai 6y ago

Say, if encryption isn't done on the device itself, then it wouldn't need to be tamper proof would it?

solarkraft 6y ago
Yeah. To my untrained eye it looks like FIPS-140 was designed for devices handling unencrypted data (anything with computation) where physical security matters and this USB stick is just a by-product of a blanket requirement that it's used.
- toyg 6y ago
  Isn't that a good thing? Anything implemented in software puts some burden on the user's side ("are you using the right OS with the right program and the right options in the right way? Just one misstep and it's all for naught"). Offloading more of that cognitive load and operational risk onto dedicated hardware is surely a good thing (if uneconomical) in practice.
  - Nullabillity 6y ago
    The USB host still has access to both the plaintext and key (once authenticated), so they both need to be fully trusted anyway. This garbage just adds yet another link to the critical chain. Oh, and now you have full plain text access over the (unencrypted and unauthenticated) USB bus. Hope you put as much effort into vetting your USB hubs!
    - toyg 6y ago
      When I was writing that post, I had actually added a second paragraph, that basically said "I hope there were some standard mechanisms for the hardware to refuse mounting without a password or some other verification system". I eventually deleted it because there might well be something of the sort already, in the bowels of TPM specs and similar. There is no reason, imho, why we shouldn't strive to "encrypt all the things", including buses, so that data is guaranteed to be available only to fully-trusted chains. This should obviously be configurable by administrators, with destruction guarantees when options are relaxed. This will not be for everyone (and yes, it will likely always be weak against the $5-wrench attack), but for people who really need it (like deployed troops, whose default opsec practices are typically too lax to be left to humans).