Darren Reed <darrenr@reed.wattle.id.au> writes:

> Woops we seem to have gotten off the rails.
> 
> 1. zero'ing out swap.

> This helps even with encrypted swap because it can't (easily) be
> retrieved for later cryptoanalysis.  I think as time goes by, old
> mechanisms which relied on residual magnetic fields be left
> polarised one way or the other will become less effect as the size
> of the magnetic area gets smaller.

There was an article in Scientific American about this (it came up on
rootprompt.org a while ago). With enough care apparently the bits can
be detected a long time after they were zeroed. If they always get
zeroed you could factor that it in when analysing the magnetic field
measurements - I'd say, write a random pattern to the swap, and the
whole swap every time, so that bit states don't get burned in (and
thus easier to detect or factor out).

[...]

> 2. encrypting swap
> I seriously doubt CPU is an issue here,

Indeed.

[...]

-- 

	http://ape.n3.net