kernel-level key management subsystem

[Bill Frantz]

travis+ml-cryptography at subspacefield.org on Monday, October 8, 2007 wrote:

>1) marking memory unpageable (avoiding swap hazard)

Encrypted swap makes the swap hazard go away, and is certainly a sweet
spot for cryptography because most of the difficult problems in key
handling "just go away".  Also, many operating systems support it.

Since keys do not have to last longer than the boot of the OS, they can
be kept in kernel memory. inaccessible to user processes, and safe from
postmortem disk analysis.  If, as most, if not all implementations do,
you have a different key for each address space, the key can go away
when the address space goes away.  You don't need to generate a key
until the first swap out for the address space, so there is time to
gather "entropy" from the timing of events such as disk I/O, network I/O
etc., making the issue of random number generation less of a problem.

For the really paranoid, keep a bit with each key saying whether it is
inverted or not, and invert it regularly to avoid memory "burn in". 
Also quickly overwrite any key-dependant data used by your symmetric
encryption algorithm.

Cheers - Bill

-----------------------------------------------------------------------
Bill Frantz        | I like the farmers' market   | Periwinkle
(408)356-8506      | because I can get fruits and | 16345 Englewood Ave
www.pwpconsult.com | vegetables without stickers. | Los Gatos, CA 95032

---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at metzdowd.com