4 ms·

GnuPG needs to push to update OpenPGP to stay alive. OpenPGP needs to be updated. It lacks modern AEAD (the practical impact of a better authentication might be

by aborsy 5y ago

GnuPG needs to push to update OpenPGP to stay alive. OpenPGP needs to be updated. It lacks modern AEAD (the practical impact of a better authentication might be low in many use cases; but it has become a political problem). The code base and CLI could also be modernized so that developers could easily use its API. Sequoia is doing a good job here. Otherwise, it risks becoming obsolete.

upofadown 5y ago
>...the practical impact of a better authentication might be low in many use cases; but it has become a political problem... In the ways that GnuPG is normally used the practical impact is zero because that is not how a stateless, offline protocol works. The content is authenticated by signing it directly, thus avoiding the extra complexity of a stateful connection oriented authentication scheme. The details here: * https://articles.59.ca/doku.php?id=pgpfan:authenticated https://articles.59.ca/doku.php?id=pgpfan:authenticated So should functionality be added here for what would be purely political reasons? Or should the political purpose be spun off into a separate utility?
- tptacek 5y ago
  This is such a random argument, and you bring it up in every thread about PGP. No, it's not at all the case that authenticated encryption is only meaningful in "stateful, online protocols", nor is it the case that signatures and authentication tags are interchangeable. Even PGP doesn't agree with this weird argument.
- phyzome 5y ago
  This is a category error. Authenticated encryption does not require statefulness. Just look at NaCl's cryptobox.
  - upofadown 5y ago
    The state is the authentication here. For stuff that is only sent once you can check the signature and you are done. Otherwise you have to maintain it for the duration of the connection and you are not going to want to sign every single message sent.
throw0101a 5y ago
> It lacks modern AEAD (the practical impact of a better authentication might be low in many use cases; but it has become a political problem). For AEAD and OpenPGP, there seem to be limitations in the message format: * https://datatracker.ietf.org/doc/html/draft-ford-openpgp-format https://datatracker.ietf.org/doc/html/draft-ford-openpgp-for... Newer versions of GPG do support AEAD modes of operation, but AFAICT they only appear in the "encrypted data" part of the resulting bucket of data and not in the "signature" part. The algorithms are generally supported though; search for 'aead': * https://www.gnupg.org/documentation/manuals/gnupg/OpenPGP-Options.html https://www.gnupg.org/documentation/manuals/gnupg/OpenPGP-Op... * https://www.gnupg.org/documentation/manuals/gcrypt/Available-cipher-modes.html https://www.gnupg.org/documentation/manuals/gcrypt/Available...
- tptacek 5y ago
  The world is going to be locked into legacy CFB+MDC PGP pretty much indefinitely (because of the need to be compatible with an installed base which, owing to asynchronous messaging, can't be queried or negotiated with). As a result, to defend PGP as an ongoing practice, you're rhetorically required to argue that authenticated encryption --- not just modern AEAD, but the whole concept of authenticating ciphertext --- is overrated; 20 years of cryptographic research be damned.
  - throw0101a 5y ago
    As per the links I posted, drafts have been written acknowledging AEAD as something worth having, and code has been added to support those cipher modes. I don't think that anyone is is ignoring authenticating cipher text. But as you said, there's an installed base of, e.g., RHEL7, that cannot handle that. Some code does not churn as fast as semi-monthly web browser releases. If you do not like PGP/GPG, you 'just' have to create a better way of doing things, and then 'just' have to convince everyone to switch their workflows to it.

by aborsy 5y ago

upofadown 5y ago

>...the practical impact of a better authentication might be low in many use cases; but it has become a political problem... In the ways that GnuPG is normally used the practical impact is zero because that is not how a stateless, offline protocol works. The content is authenticated by signing it directly, thus avoiding the extra complexity of a stateful connection oriented authentication scheme. The details here: * https://articles.59.ca/doku.php?id=pgpfan:authenticated https://articles.59.ca/doku.php?id=pgpfan:authenticated So should functionality be added here for what would be purely political reasons? Or should the political purpose be spun off into a separate utility?

tptacek 5y ago
This is such a random argument, and you bring it up in every thread about PGP. No, it's not at all the case that authenticated encryption is only meaningful in "stateful, online protocols", nor is it the case that signatures and authentication tags are interchangeable. Even PGP doesn't agree with this weird argument.
phyzome 5y ago
This is a category error. Authenticated encryption does not require statefulness. Just look at NaCl's cryptobox.
- upofadown 5y ago
  The state is the authentication here. For stuff that is only sent once you can check the signature and you are done. Otherwise you have to maintain it for the duration of the connection and you are not going to want to sign every single message sent.

throw0101a 5y ago

> It lacks modern AEAD (the practical impact of a better authentication might be low in many use cases; but it has become a political problem). For AEAD and OpenPGP, there seem to be limitations in the message format: * https://datatracker.ietf.org/doc/html/draft-ford-openpgp-format https://datatracker.ietf.org/doc/html/draft-ford-openpgp-for... Newer versions of GPG do support AEAD modes of operation, but AFAICT they only appear in the "encrypted data" part of the resulting bucket of data and not in the "signature" part. The algorithms are generally supported though; search for 'aead': * https://www.gnupg.org/documentation/manuals/gnupg/OpenPGP-Options.html https://www.gnupg.org/documentation/manuals/gnupg/OpenPGP-Op... * https://www.gnupg.org/documentation/manuals/gcrypt/Available-cipher-modes.html https://www.gnupg.org/documentation/manuals/gcrypt/Available...

tptacek 5y ago
The world is going to be locked into legacy CFB+MDC PGP pretty much indefinitely (because of the need to be compatible with an installed base which, owing to asynchronous messaging, can't be queried or negotiated with). As a result, to defend PGP as an ongoing practice, you're rhetorically required to argue that authenticated encryption --- not just modern AEAD, but the whole concept of authenticating ciphertext --- is overrated; 20 years of cryptographic research be damned.
- throw0101a 5y ago
  As per the links I posted, drafts have been written acknowledging AEAD as something worth having, and code has been added to support those cipher modes. I don't think that anyone is is ignoring authenticating cipher text. But as you said, there's an installed base of, e.g., RHEL7, that cannot handle that. Some code does not churn as fast as semi-monthly web browser releases. If you do not like PGP/GPG, you 'just' have to create a better way of doing things, and then 'just' have to convince everyone to switch their workflows to it.