Data recovery companies ought to have the integrity to just say no to a job, if they cannot do it risk free. Trying and failing with the risk of damaging the original data could be very costly to the customer, even if they don't charge money - the customer's lost data could be priceless.
The challenges will be different. Flash loses its charge in 30 years, most disks are encrypted, and on-site physical backups are mostly a thing of the past. The source might survive in a cloud repo, but it'll either be tied up for legal reasons or deleted when the customer stops paying the bill. But storage is cheap and getting cheaper!
Encryption for 30 years ago? Trivially breakable with quantum
I wouldn't be so sure - quantum computers aren't nearly as effective for symmetric algorithms as they are for pre-quantum asymmetric algorithms.
I can't think of anything from 30 years ago that isn't just a joke today. The same will likely be true by 2050, quantum computing or not. I wonder how many people realise this?
Even if one disagrees with my certainty, I think people should still plan for the concept that there's a strong probability it will be so. Encryption is really not about preventing data exposure, but about delaying it.
Any other view regarding encryption means disappointment.
You are underestimating the exponential possibilities of keys.
>plus general vulnerabilities and weaknesses discovered, will ensure that anything encrypted today is easily readable in the future.
You can't just assume that there is always going to be new vulnerabilities that cause it to be broken. It ignores that people have improved at designing secure cryptography over time.
An example being, destroying sensitive backup media upon its retirement, regardless of data encryption.
AES is only 3 years shy of 30.
If you used MD5 as a keystream generator I believe that would still be secure and that's 33 years old.
3DES is still pretty secure, isn't it? That's 44 years old.
As for today's data, there's always risk into the future but we've gotten better as making secure algorithms over time and avoiding quantum attacks seems to mostly be a matter of doubling key length. I'd worry more about plain old leaks.
How do you tell which will fall, and which will succeed in 30 years?
All this said, I just think proper mental framing helps. Considering the value of encrypted data, in 30 years, if it is broken.
In many cases... who cares. In others, it could be unpleasant.
There are a lot of interactive systems that have attacks on their key exchange or authentication. And there are hashes that have collision attacks.
But compromises that let you figure out a key that's no longer in use have not been common for a while. And even md5 can't be reversed.
I agree with you about being wary, but I think encryption itself can be one of the stronger links in the chain, even going out 30 years.
The gold standard 30 years ago was PGP. RSA 1024 or 2048 for key exchange. IDEA symmetric cipher.
This combination is, as far as I am aware, still practically cryptographically secure. Though maybe not in another 10 or 20 years. (RSA 1024 is not that far from brute forcing with classical machines.)
In summary, it estimates the cost at $3.5 billion using commodity hardware, and I'd expect a purpose-built system could bring that cost down by an order of magnitude.
The reality is, as soon as humanity figures out how to distinguish between two values (magnetic flux, voltage, pits/lands, etc) we use it to store more data, or move it faster.
The end.
On top of that static wear leveling can move all your data around when your disk is idle, and TRIM will effectively zero your unused areas when you are not looking.
So, it's a very different landscape.
I enjoyed skimming through this: https://github.com/HighwayFrogs/frogger2-vss/blob/main/teamS...
Take this!
https://github.com/Kneesnap/onstream-data-recovery/blob/main...
My takeaway is that you should choose a passion project as your hobby and put in the time to learn and do whatever is necessary to achieve your goal on your own or together with similarly motivated people rather than relying on anyone external you have to pay - things go downhill fairly often and quickly it seems. Is any business a scam to some degree nowadays?
Tape storage can be an absolute nightmare. Most will do the writes, some will say they verify with a read, but few actually test with a full restore. Just because the software says it can read the tape to show you the listings does not mean it can read the files themselves. This was alluded to in TFA(TFV??) but been there done that on trying to read from a bum tape/bad write. It gets worse if you write in one tape drive and read from another also mentioned in TFV. Now I feel old just thinking about it all
I've recovered data from media a number of times a recovery company said it was unrecoverable with no particular difficulty.
Pretty much all tape backup software writes headers as it is streaming the file to tape. Just more bytes in the buffer.
For normal restores it consults its local database because that is way faster. If you don't have the local database you do a "Catalog Tape" operation that scans the file headers on the tape to reconstruct the database. For whatever reason ARCServe couldn't complete the catalog with that specific kind of tape. Whether that was the specific version he found or was a general problem with support for those tape drives I don't know.
NaOH•7mo ago
The long road to recover Frogger 2 source from tape drives - https://news.ycombinator.com/item?id=36061574 - May 2023 (213 comments)