Timeline for FileVault security hole when used on SSDs
Current License: CC BY-SA 3.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 22, 2016 at 4:12 | vote | accept | David Braun | ||
| Feb 20, 2016 at 10:07 | history | tweeted | twitter.com/askdifferent/status/700985077317169152 | ||
| Feb 19, 2016 at 17:20 | answer | added | bmike♦ | timeline score: 3 | |
| May 25, 2015 at 11:41 | comment | added | n1000 | Apple kind of acknowledges this: support.apple.com/en-us/HT201949 Also interessting: discussions.apple.com/docs/DOC-5521 | |
| May 25, 2015 at 8:30 | comment | added | user125447 | I would also skip the original encryption prompt and enable FileVault after the machine is configured. Then transfer your data over to the newly encrypted drive. As far as when you want to sell the computer later, you can just format the drive using disk utility and then install a fresh copy of the OS on the drive. Even though you can't securely erase files on an SSD the same, if you format the encrypted partition when you sell it, all of your files would still be encrypted and inaccessible to the new owner. | |
| Mar 1, 2014 at 13:16 | history | edited | David Braun | CC BY-SA 3.0 |
Add zigg's suggestion to increase entropy before enabling FileVault.
|
| Mar 1, 2014 at 8:00 | answer | added | user71659 | timeline score: -1 | |
| Feb 28, 2014 at 22:47 | answer | added | taddy hoops | timeline score: -1 | |
| Feb 28, 2014 at 21:22 | comment | added | David Braun | @zigg Your speculation is spot on! See page 9 of eprint.iacr.org/2012/374.pdf: "For highly security-critical scenarios, the PRNG should be reseeded by manually writing entropy to /dev/random before activation of FileVault 2." | |
| Feb 28, 2014 at 20:36 | comment | added | Mattie | @DavidBraun Unfortunately, no. All I can do is speculate based on how I know it's needed. 😒 It is possible that entropy is collected during installation, though… | |
| Feb 28, 2014 at 20:31 | comment | added | David Braun | @zigg Do you have any evidence that FileVault is using only as much entropy as is available, as opposed to (for example) GnuPG's method of waiting for enough entropy before making a key? | |
| Feb 28, 2014 at 20:29 | comment | added | David Braun | @dwightk I don't know the implementation details of Apple's Secure Empty Trash but I've learned from the articles linked above that algorithms which worked on HDD by overwriting data no longer work on SSD. The linked support page implies that encryption before writing is now the only viable solution. | |
| Feb 28, 2014 at 18:17 | comment | added | Mattie | @da4 Do you doubt the FileVault implementation? The underlying encryption? Or users using it properly? As to the OP, I think your strategy is reasonable. I'd also hold off on FV for a bit (maybe until updates are applied, etc.) on a brand-new system to give entropy a chance to build and make the key stronger. | |
| Feb 28, 2014 at 16:29 | comment | added | da4 | Anyone who has physical control of a computer will ultimately be able to recover some data from it - it's simply a question of how much effort and time would be required in order to recover how much data. That said, if you're worried about any data ever being "in the clear," then yes, I agree with your solution - I'd let FV finish and then migrate or manually copy. | |
| Feb 28, 2014 at 16:14 | comment | added | dwightk | Are you sure that the manner in which Apple securely emptied trash in early 2011 (Snow Leopard was the latest OS X) is how they do it now? Also, are you sure when the SSD is deleted using other means than securely emptying trash the underlying mechanism is the same? | |
| Feb 28, 2014 at 15:38 | history | asked | David Braun | CC BY-SA 3.0 |