Re: Passwords & Passphrases

[Gary Flynn <flynngn () JMU EDU>]

Steven Carmody wrote:

> The US InCommon Federation ( http://www.incommonfederation.org/ ) is
> working with Federal government agencies to provide campuses with
> federated access to agency websites. Recently, NIH joined InCommon (and
> will be making its grants management website available in the coming
> months). IC is also talking to other agencies.
>
> In order to access federal sites at e-authentication level 2 (eg NIH
> grants mgmt), campus security practices will have to meet a number of
> criteria. One of these criteria is:
>
> > Strong resistance to guessing shared secret
>
> > 1. The PIN (numeric-only) or password, and the controls used to limit
> > on-line guessing attacks shall ensure that an attack targeted against
> > a selected user's PIN or password shall have a probability of success
> > of less than 2-14 (1 chance in 16,384) over the life of the PIN or
> > Password.  2. The PIN (numeric-only) or password shall have at least
> > 10 bits of min-entropy (a measure of the difficulty that an attacker
> > faces to guess the most commonly chosen password used in a system) to
> > protect against untargeted attack.
> > Refer to NIST SP 800-63 Appendix A, and the NIST Shared Secret Entropy
> > Spreadsheet to calculate resistance to online guessing.
>
> So, if a campus allows a user to change their password, the new password
> must be "at least" as strong as these requirements. While this might not
> constitute "best practice", it does set a minimum bar.
>
> The entropy spreadsheet can be found here:
>
> http://www.cio.gov/eauthentication/credential_suite.cfm


Note that it appears to require a password guessing lockout.

If you open the spreadsheet and go to the password calculation tab,
then to the Password Management Rules section, you'll see the first
value is 'Lockout Mechanism'. If you set that to none, regardless
of other settings, you'll get a 'SYSTEM FAILS TEST' status with the
explanation "No mechanism to prevent attack from exceeding limit of
1,024 tries". Or taken to extreme with a 40 character password and
all complexity rules enabled, "No mechanism to prevent attacker
from exceeding limit of 36,028,797,018,964,000 tries".



--
Gary Flynn
Security Engineer
James Madison University
www.jmu.edu/computing/security

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature