Some combination has to be enough. Because otherwise you could use that property (plus error resistance coding) to store arbitrary amounts of data on your hard drive.
If you are able to get at the entire physical media then a certain combination of high and low bit writing is necessary or one can read some previous bits with specialized hardware. There is a DoD standard for writing enough combinations.
But with modern drives you are not getting at the physical layout. The drive may detect an error and remap surrounding data for you. Consequently, some data surrounding some badblock(s) in the past will potentially never be overwritten by your process. Someone with separate hardware or an alternate firmware could always retrieve it.
I guess the right solution would be to store your data only in encrypted form on your drive. Then you will only have to worry about safely deleting the key.
The size of the error resistance coding you'll need will quickly dominate the size of your data. Also the argument is a probabilistic one. It's not that certain data will be recoverable, only that there may be a non-zero probability that some data may be recoverable.
Comments
Some combination has to be enough. Because otherwise you could use that property (plus error resistance coding) to store arbitrary amounts of data on your hard drive.
If you are able to get at the entire physical media then a certain combination of high and low bit writing is necessary or one can read some previous bits with specialized hardware. There is a DoD standard for writing enough combinations.
But with modern drives you are not getting at the physical layout. The drive may detect an error and remap surrounding data for you. Consequently, some data surrounding some badblock(s) in the past will potentially never be overwritten by your process. Someone with separate hardware or an alternate firmware could always retrieve it.
Yes, the badblocks argument is a good one.
I guess the right solution would be to store your data only in encrypted form on your drive. Then you will only have to worry about safely deleting the key.
The size of the error resistance coding you'll need will quickly dominate the size of your data. Also the argument is a probabilistic one. It's not that certain data will be recoverable, only that there may be a non-zero probability that some data may be recoverable.