In reply to Rob Parsons:
>The Guttman-style scare stories rely on probabilistic analysis using very sophisticated hardware tools. So, maybe you can successfully recover the odd bit (though you won't be *sure* you've done it) - but, recovering any significant amount of data? I don't believe it.
What I have heard is similar to what you have written above.
Some of the magnetic domains on the stepper track (the one that guides the heads), get magnetised when you write data. As data is encoded with error correction, the recovery of that data is easier than if it didn't have have a CRC. Scanning electron microscopes are needed to detect this. Which means the adversary only needs access to kit a reasonable university may have kicking about.
However... Not all disks will bleed data. As data densities have increased, the accuracy needed to write the data has increased as well. You cannot be certain when any bleed data was written. So it could be years or days old.
In a previous role, I used Dban followed by a pillar drill. Any adversary would have to be very well funded. They would probably opt for the far cheaper, quicker and more accurate option of bunging someone a few grand in a brown envelope.
Disk platters make surprisingly good wind chimes.