5 ms·

> I used the parameters suggested in the documentation [1], which follow the RFC Where in the RFC is your parameter set mentioned? I don't see it anywhere[0].

by d-z-m 2y ago

> I used the parameters suggested in the documentation [1], which follow the RFC Where in the RFC is your parameter set mentioned? I don't see it anywhere[0]. The only parameter set I see mentioned with memory requirement as low as 64 MiB have t=3, not t=1. I believe the Go documentation might be outdated. [0]: https://datatracker.ietf.org/doc/rfc9106/ https://datatracker.ietf.org/doc/rfc9106/ > I agree pepper confusing (because the pepper is supposedly secret), but this is not a salt either, as a salt is supposed to be public, but unique. Do you have better naming that you can suggest? If it's tweakable at build, perhaps pepper makes more sense.

ncruces 2y ago
Yep, I just made it tweakable at build, which was always the intent, although I expect the default to be popular. https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adiantum/adiantum.go#L14 https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adia... That's unfortunate about the default parameters, but note that you can also replace the KDF altogether (besides just not using it). You just need to implement this interface, with any HBSH construction and KDF: https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adiantum/api.go#L54 https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adia... If you keep the HBSH and change the KDF, your file format will be “compatible.”
- d-z-m 2y ago
  > That's unfortunate about the default parameters, but note that you can also replace the KDF altogether (besides just not using it). Right...I would argue if you're going to expose it to users as an option, your parameter set should meet the minimum hardness parameter set in the RFC. Yours is falling off the bottom.
  - ncruces 2y ago
    I don't disagree, and after some consideration and testing, will use the RFC recommendation of 3 iterations, 4 lanes, 64 MiB RAM. However, it's also important to take note of how the KDF is used, and what is used by comparable solutions in equivalent settings. For instance, the official extension, sold by the SQLite authors, uses RC4: “For the "-textkey" option, up to 256 bytes of the passphrase are hashed using RC4 and the hash value becomes the encryption key. Note that in this context the RC4 algorithm is being used as a hash function, not as a cryptographic function, so the fact that RC4 is a cryptographically weak algorithm is irrelevant.” The textkey= parameter is not really appropriate for interactive use, rather it is a convenient way for a user to specify a key in text form, as a URI parameter, without weakening it (by e.g. making NULs impossible, and reserved characters inconvenient to encode). If you're taking a password from a user interactively, you should really be using your own password hashing, then using hexkey=. I'll try to make this clearer in the documentation.
    - d-z-m 2y ago
      > However, it's also important to take note of how the KDF is used Agree completely. I'm glad that t=3 works for your usage requirements. Was hoping it wouldn't introduce an intolerable amount of latency. > and what is used by comparable solutions in equivalent settings. I interpreted the above and the linked documentation snippet as trying to say: "Yes, the KDF was sub-par, but alternatives are worse". Is this what you were trying to convey? The fact that the official encryption extension is using sub-par KDF(the sentence about it being a hash and not a "cryptographic" function is insane, btw) has no bearing on your project or its implementation.
      - ncruces 2y ago
        It is and it isn't. Can I plead the 5th? Yes, the fact that they're using crappy encryption doesn't excuse me. But (IMO) context matters, and I can't really use something that causes opening a connection to take 1s on a low(er) end server. That would be appropriate for interactive use (i.e. ask a human user for a password), but not every time a connection drops and your server needs another one. And I'm weary to add caching at the library level and risk exposing a side channel.
        d-z-m 2y ago
        > And I'm weary to add caching at the library level and risk exposing a side channel. You mean a cache-timing attack against adiantum?
        ncruces 2y ago
        I mean, caching the KDF result at the library level, without leaking that (e.g.) two databases share a key (that's the first one that came to mind, and why I removed the feature, instead of plugging the hole just to find another one).

by d-z-m 2y ago

ncruces 2y ago

Yep, I just made it tweakable at build, which was always the intent, although I expect the default to be popular. https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adiantum/adiantum.go#L14 https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adia... That's unfortunate about the default parameters, but note that you can also replace the KDF altogether (besides just not using it). You just need to implement this interface, with any HBSH construction and KDF: https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adiantum/api.go#L54 https://github.com/ncruces/go-sqlite3/blob/67d859a5/vfs/adia... If you keep the HBSH and change the KDF, your file format will be “compatible.”

d-z-m 2y ago
> That's unfortunate about the default parameters, but note that you can also replace the KDF altogether (besides just not using it). Right...I would argue if you're going to expose it to users as an option, your parameter set should meet the minimum hardness parameter set in the RFC. Yours is falling off the bottom.
- ncruces 2y ago
  I don't disagree, and after some consideration and testing, will use the RFC recommendation of 3 iterations, 4 lanes, 64 MiB RAM. However, it's also important to take note of how the KDF is used, and what is used by comparable solutions in equivalent settings. For instance, the official extension, sold by the SQLite authors, uses RC4: “For the "-textkey" option, up to 256 bytes of the passphrase are hashed using RC4 and the hash value becomes the encryption key. Note that in this context the RC4 algorithm is being used as a hash function, not as a cryptographic function, so the fact that RC4 is a cryptographically weak algorithm is irrelevant.” The textkey= parameter is not really appropriate for interactive use, rather it is a convenient way for a user to specify a key in text form, as a URI parameter, without weakening it (by e.g. making NULs impossible, and reserved characters inconvenient to encode). If you're taking a password from a user interactively, you should really be using your own password hashing, then using hexkey=. I'll try to make this clearer in the documentation.
  - d-z-m 2y ago
    > However, it's also important to take note of how the KDF is used Agree completely. I'm glad that t=3 works for your usage requirements. Was hoping it wouldn't introduce an intolerable amount of latency. > and what is used by comparable solutions in equivalent settings. I interpreted the above and the linked documentation snippet as trying to say: "Yes, the KDF was sub-par, but alternatives are worse". Is this what you were trying to convey? The fact that the official encryption extension is using sub-par KDF(the sentence about it being a hash and not a "cryptographic" function is insane, btw) has no bearing on your project or its implementation.
    - ncruces 2y ago
      It is and it isn't. Can I plead the 5th? Yes, the fact that they're using crappy encryption doesn't excuse me. But (IMO) context matters, and I can't really use something that causes opening a connection to take 1s on a low(er) end server. That would be appropriate for interactive use (i.e. ask a human user for a password), but not every time a connection drops and your server needs another one. And I'm weary to add caching at the library level and risk exposing a side channel.
      - d-z-m 2y ago
        > And I'm weary to add caching at the library level and risk exposing a side channel. You mean a cache-timing attack against adiantum?
        ncruces 2y ago
        I mean, caching the KDF result at the library level, without leaking that (e.g.) two databases share a key (that's the first one that came to mind, and why I removed the feature, instead of plugging the hole just to find another one).