SD Card Data Recovery

[Armin]

Doesn't sound like you had a 'professional data recovery' company.

Based on my personal experience, they are not professionals. Though they have managed to build a data-recovery-empire in Arizona with over 37 store locations, probably by being professional "over 95%" of the time. My experience is apparently in the minority.

[Rule]

Doesn't sound like you had a 'professional data recovery' company.

Based on my personal experience, they are not professionals. Though they have managed to build a data-recovery-empire in Arizona with over 37 store locations, probably by being professional "over 95%" of the time. My experience is apparently in the minority.

Maybe their readily available recovery software works for 95% of the problems people bring into them...

[Armin]

That's likely as well. I gave them the benefit of any doubt.

Anyways, since I imagine these types of formatting mistakes happen more often than it should, this experience had me thinking about a possible solution for piecing together fragmented media files from a formatted drive.

The proposed program would piece back together the fragmented information with the following method:

• Based on my limited knowledge, I've assumed that a media file has a basic structure and syntax for it to be read and executed properly. For the sake of this thought experiment, I will call any file with this basic structure and syntax a 'solution'.
• The program would calculate all possible 'solutions' for the array of fragmented files.
  The more fragments, the more possible 'solutions'.
• (A more intelligent program could perhaps calculate another, smaller number of 'likely' solutions.)
• Depending on the user-input, the program would then render either all possible solutions, or all 'likely' solutions.
• The user would then sort through the solutions, and choose the correct solution.

Obviously, the program would require a specific algorithm for each codec, of each file-type.


Let me know if I was unclear in anyway.

[Sidoh]

My understanding of the problem with data loss leads me to believe this sort of thing isn't really possible. The solutions vector would basically include any sequence of blocks that aren't accounted for?

Perhaps there's some sort of sequencing data you could search for, and that's probably what you're getting at. I don't know enough about this stuff to be more helpful.

[iago]

Compressed data is essentially random, so it'd be super difficult to figure out where blocks start/end and how they go together.

[Sidoh]

Ah, sure. Maximized entropy and all that.