phc-discussions - Re: [PHC] Some KDF stumbling blocks, plus Common "memory-hard" approaches and amortized attack costs.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAMtf1HvqeuzNZu3EDJQ=oeaVzScrTEbjDJxcBxUusAok4pSBSg@mail.gmail.com>
Date: Sat, 13 Dec 2014 09:11:59 +0800
From: Ben Harris <ben@...rr.is>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Some KDF stumbling blocks, plus Common "memory-hard"
 approaches and amortized attack costs.

On 13 December 2014 at 06:38, Gregory Maxwell <gmaxwell@...il.com> wrote:

> A third is that some applications must support small memory embedded
> devices. For example, hardware wallets with 128kbytes ram. Cost, power
> usage, durability against sidechannels, etc. all contribute to memory
> limitations on these devices. Additional ram is not an option, the
> option is to instead use a malware infested desktop PC or smartphone
> or not use a KDF.
>
> These devices are never themselves going to support KDFs with compute
> or memory hardness beyond snake-oil levels.  I am saddened that none
> of the proposals supported delegation scheme with
> information-theoretic security ( which are possible,
> https://bitcointalk.org/index.php?topic=311000.0 ) but not surprised
> since I was unable to come up for a construction for one which
> achieved the memory hardness properties which are currently
> fashionable for KDFs.   Even lacking that, any of these devices will
> need a delegation scheme of some kind. If one isn't supported they
> will either reject the strong KDF or achieve one informally via
> pre-processing.  e.g. H1||H2 = HMAC(salt,password)   Key =
> H(StrongKDF(H1,user)||H2) .   It would be be preferable if such
> delegation were specified explicitly rather than applications rolling
> their own incompatible and potentially less secure versions.
>

Thanks for the well developed question exploring a variety of areas. For
the embedded side, the option is to have the PasswordBasedKDF either on the
device or off the device. From what I have looked at, the common solution
is to securely store the output of the PBKDF on the device and just have
the device use it in a KDF to generate the output keys. Some entries, like
Catena (https://password-hashing.net/submissions/specs/Catena-v2.pdf),
include the second part in their specs (page 48).

Handling the PBKDF on the device might be the ideal situation, but as you
say it is a harder problem.

Content of type "text/html" skipped