This is why our backups at work write to a storage bucket with permissions such that they can create new files but not delete old ones. I'd definitely recommend this approach to everyone who can afford the storage space.
This is why our backups at work write to a storage bucket with permissions such that they can create new files but not delete old ones. I'd definitely recommend this approach to everyone who can afford the storage space.
For anyone who has serious (I.e. $$$) need of that they already have tapes and optical WORM media though.
You can do something conceptually similar with any sort of NAS that provides immutable snapshots as long as the management and control is effectively out of band.
The out of band part is the key. Our SAN data has snapshots. The backups are written to another storage device that only has an API key to write them to B2 storage. An attacker would effectively need to completely compromise multiple admins in the organization to get at all the stages of data duplication, and frankly there is no additional line of defense for total compromise if the attacker is willing to wait for physical tape or disk swaps.
Fortunately for ransomware, time is money for them too.
Sigh. When are people going to accept that software switches are inherently not secure? How many times must these fail?
> It's difficult to do that unless the filesystem has append only functionality, metadata blocks are rewritten all the time even if data isn't.
There's no reason to continue writing anything to a hard drive once the backup to it is finished.
Even more secure would be hardware write only storage. CD-ROMs fit in this category, but they aren't big enough.
But all we need are hard disk drives with a physical write-enable switch. Turn it on, write your backup, turn it off. No software can then alter it.
A stupidly simple idea, and yet every time I mention it in HN it gets dismissed, denigrated, etc. Apparently people like malware, ransomware, etc. :-(
A physical read-only switch is required.
Not ironclad but pretty good. Issues with it that first come to mind are the live operating system image was already compromised when it was written to say the USB disk, or compromised firmware, and of course user error (nothing to do with ransomware in this case). I am not familiar with this stuff so I may be missing something very important, if so tell me about it.
Just think of all the security issues that would just go away with physical write-enable switches.
Heck, I'd go further, and demand from disk makers a physical write enable switch for a separate volume. Use that volume for the system software.