|
Top-level
 @nixCraft EDIT: Been answered. I would imagine even if they had hit the board if they miss the chip itself they risk that some expert could recover data from it anyway. I don't know much about it, but isn't the old fashioned write 10x or so still effective even on SSDs? I know things get remapped and wear cycling even has reserved parts, but enough writes of random data should fully clear it eventually, right? (Though not sure if 10x is right or not with SSDs. That's from magnetic media.) 20 comments
@4censord @nixCraft I know about erasing blocks, yeah. But I presume we're talking about extreme measures when someone goes as far as to actually drill holes in it. In theory it's possible to recover old data from erased cells -- or at least so I've heard. Presumably the idea here is to ensure that is no longer possible. Sure is wasteful as heck to drill it IMO.  @nazokiyoubinbou @nixCraft the secure erase thing uses encryption to prevent data recovery. https://wiki.archlinux.org/title/Self-encrypting_drives @nazokiyoubinbou @4censord @nixCraft Technically, TRIM erases pages. Pages on a flash chip can contain several blocks. Erasing a single flash page is basically the voltage spike, but to only the one page and for a shorter time. The chip-level erase lasts longer in part to make sure everything gets fully saturated. Different trigger and different scale of effect, but it’s the same mechanism.  @4censord  @4censord @nazokiyoubinbou @nixCraft This only works on SSD drives that supports this feature. @nazokiyoubinbou @nixCraft well, the reserved parts oft the SSD are often not accessible via the normal SSD data controller. So rewrite often works not 100%.  @nazokiyoubinbou @nixCraft Maybe I’m missing something, but with SSDs it’s just better to encrypt the content of the drive. Destroying the key then effectively erases the data. @melgu @nixCraft I mean someone going extreme enough to drill through them is presumably assuming that an attacker might be able to extract data, so I would assume they would not trust encryption to be absolute either. (There are also other side effects like performance loss if it uses a complex enough encryption to be worthwhile.) Either way though, a full cell erase should be sufficient. No need to be super wasteful and destructive. @nazokiyoubinbou @nixCraft Isn’t drive encryption / decryption done in hardware these days anyways? @nazokiyoubinbou @nixCraft I did? I was only talking about the performance loss due to the encryption. All your points are valid, though. @melgu @nazokiyoubinbou @nixCraft I don't see that flushing the key is particularly future proof. The data remains only whilst the encryption remains strong. @nazokiyoubinbou @nixCraft Overwriting 10x is only sensible for (very) old hard disks or floppy disks with wide tracks. Current hard drives pack tracks so close that writing one may impact neighbouring tracks. So overwritung the complete disk once (e.g. 'dd bs=1M </dev/zero >/dev/sdX' on linux) will do the trick. On SSDs, due to wear leveling algorithms parts of data may survive overwriting in spare sectors but reading date from there will be hard. @nazokiyoubinbou Full disk encryption with e.g. LUKS also is a good first line of defense. @nazokiyoubinbou @nixCraft Use disk encryption from the start and without the key all data on the disk looks random. |
Gregory's Server
@nazokiyoubinbou @nixCraft there is a point of overwrites where every flash cell has been changed. But, that costs loads of disk wear. Instead, one can issue a data secure erase command to drop all data, that even restores some write performance