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  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]
Date: Tue, 02 Sep 2014 08:11:13 -0400
From: Bill Cox <waywardgeek@...hershed.org>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] A review per day - Schvrch

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Here's how Solar Designer would PWN any machine asking for a password
that uses the current version of Schvrch.  Here's the top of the PHS
function:

int PHS(void *out, size_t outlen,
        const void *in, size_t inlen,
        const void *salt, size_t saltlen,
        unsigned int t_cost, unsigned m_cost)
{
    int statelen = 256, j;
    uint64_t state[256] = {0};
    uint64_t memcost;
    memcost = (m_cost + 1) * statelen;
    uint64_t * memstate;
    uint64_t rounds = 4, i;

    memmove(&state[0], in, inlen);

There is *no* input checking.  As I said for POMELO, this *must* be
fixed.  Without it, we simply pass inlen > 256, and a custom made
password that has a return instruction encoded in it.  We overwrite
the return address with our own value embeded in the password, and
when this function returns, it returns to any known address we like.
In Linux, this would typically be some function in libc, such as the
"system" command, which can execute arbitrary commands.

I do believe that basic input checking should be part of the reference
design.  It tells us what valid parameters are, which impacts
cryptanalysis.  It also reduces the chance of mistakes like this
escaping into the wild.

I know that there is a guy doing python wrappers for PHC entries using
the PHS function.  If he puts that live on a server anywhere, we can
own that server.  POMELO has this problem, too.  I do not consider
fixing this optional.

For such a small program, there are a lot of attack opportunities.
When I read in the paper that the author considers it secure because
it retains the password entropy, I stopped reading.  There may be some
very cool math there, but the implementation needs a lot work.

Until these devastating attacks are fixed, I consider Schvrch
dangerous to use.

Bill
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
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=tRKh
-----END PGP SIGNATURE-----

Powered by blists - more mailing lists