8 6. Backup and Data Recovery
9 7. Interoperability with other Disk Encryption Tools
10 8. Issues with Specific Versions of cryptsetup
11 9. References and Further Reading
20 This is the FAQ (Frequently Asked Questions) for cryptsetup. It
21 covers Linux disk encryption with plain dm-crypt (one passphrase,
22 no management, no metadata on disk) and LUKS (multiple user keys
23 with one master key, anti-forensic features, metadata block at
24 start of device, ...). The latest version of this FAQ should
25 usually be available at
26 https://gitlab.com/cryptsetup/cryptsetup/wikis/FrequentlyAskedQuestions
31 ATTENTION: If you are going to read just one thing, make it the
32 section on Backup and Data Recovery. By far the most questions on
33 the cryptsetup mailing list are from people that managed to damage
34 the start of their LUKS partitions, i.e. the LUKS header. In
35 most cases, there is nothing that can be done to help these poor
36 souls recover their data. Make sure you understand the problem and
37 limitations imposed by the LUKS security model BEFORE you face
38 such a disaster! In particular, make sure you have a current header
39 backup before doing any potentially dangerous operations.
41 SSDs/FLASH DRIVES: SSDs and Flash are different. Currently it is
42 unclear how to get LUKS or plain dm-crypt to run on them with the
43 full set of security features intact. This may or may not be a
44 problem, depending on the attacker model. See Section 5.19.
46 BACKUP: Yes, encrypted disks die, just as normal ones do. A full
47 backup is mandatory, see Section "6. Backup and Data Recovery" on
48 options for doing encrypted backup.
50 CLONING/IMAGING: If you clone or image a LUKS container, you make a
51 copy of the LUKS header and the master key will stay the same!
52 That means that if you distribute an image to several machines, the
53 same master key will be used on all of them, regardless of whether
54 you change the passphrases. Do NOT do this! If you do, a root-user
55 on any of the machines with a mapped (decrypted) container or a
56 passphrase on that machine can decrypt all other copies, breaking
57 security. See also Item 6.15.
59 DISTRIBUTION INSTALLERS: Some distribution installers offer to
60 create LUKS containers in a way that can be mistaken as activation
61 of an existing container. Creating a new LUKS container on top of
62 an existing one leads to permanent, complete and irreversible data
63 loss. It is strongly recommended to only use distribution
64 installers after a complete backup of all LUKS containers has been
67 UBUNTU INSTALLER: In particular the Ubuntu installer seems to be
68 quite willing to kill LUKS containers in several different ways.
69 Those responsible at Ubuntu seem not to care very much (it is very
70 easy to recognize a LUKS container), so treat the process of
71 installing Ubuntu as a severe hazard to any LUKS container you may
74 NO WARNING ON NON-INTERACTIVE FORMAT: If you feed cryptsetup from
75 STDIN (e.g. via GnuPG) on LUKS format, it does not give you the
76 warning that you are about to format (and e.g. will lose any
77 pre-existing LUKS container on the target), as it assumes it is
78 used from a script. In this scenario, the responsibility for
79 warning the user and possibly checking for an existing LUKS header
80 is shifted to the script. This is a more general form of the
83 LUKS PASSPHRASE IS NOT THE MASTER KEY: The LUKS passphrase is not
84 used in deriving the master key. It is used in decrypting a master
85 key that is randomly selected on header creation. This means that
86 if you create a new LUKS header on top of an old one with
87 exactly the same parameters and exactly the same passphrase as the
88 old one, it will still have a different master key and your data
89 will be permanently lost.
91 PASSPHRASE CHARACTER SET: Some people have had difficulties with
92 this when upgrading distributions. It is highly advisable to only
93 use the 95 printable characters from the first 128 characters of
94 the ASCII table, as they will always have the same binary
95 representation. Other characters may have different encoding
96 depending on system configuration and your passphrase will not
97 work with a different encoding. A table of the standardized first
98 128 ASCII characters can, e.g. be found on
99 http://en.wikipedia.org/wiki/ASCII
102 * 1.3 System specific warnings
104 - Ubuntu as of 4/2011: It seems the installer offers to create
105 LUKS partitions in a way that several people mistook for an offer
106 to activate their existing LUKS partition. The installer gives no
107 or an inadequate warning and will destroy your old LUKS header,
108 causing permanent data loss. See also the section on Backup and
111 This issue has been acknowledged by the Ubuntu dev team, see here:
112 http://launchpad.net/bugs/420080
114 Update 4/2013: I am still unsure whether this has been fixed by
115 now, best be careful. They also seem to have added even more LUKS
116 killer functionality to the Ubuntu installer. I can only strongly
117 recommended to not install Ubuntu on a system with existing LUKS
118 containers without complete backups.
121 * 1.4 My LUKS-device is broken! Help!
123 First: Do not panic! In many cases the data is still recoverable.
124 Do not do anything hasty! Steps:
126 - Take some deep breaths. Maybe add some relaxing music. This may
127 sound funny, but I am completely serious. Often, critical damage is
128 done only after the initial problem.
130 - Do not reboot. The keys mays still be in the kernel if the device
133 - Make sure others do not reboot the system.
135 - Do not write to your disk without a clear understanding why this
136 will not make matters worse. Do a sector-level backup before any
137 writes. Often you do not need to write at all to get enough access
138 to make a backup of the data.
142 - Read section 6 of this FAQ.
144 - Ask on the mailing-list if you need more help.
147 * 1.5 Who wrote this?
149 Current FAQ maintainer is Arno Wagner <arno@wagner.name>. If you
150 want to send me encrypted email, my current PGP key is DSA key
151 CB5D9718, fingerprint 12D6 C03B 1B30 33BB 13CF B774 E35C 5FA1 CB5D
154 Other contributors are listed at the end. If you want to contribute,
155 send your article, including a descriptive headline, to the
156 maintainer, or the dm-crypt mailing list with something like "FAQ
157 ..." in the subject. You can also send more raw information and
158 have me write the section. Please note that by contributing to this
159 FAQ, you accept the license described below.
161 This work is under the "Attribution-Share Alike 3.0 Unported"
162 license, which means distribution is unlimited, you may create
163 derived works, but attributions to original authors and this
164 license statement must be retained and the derived work must be
165 under the same license. See
166 http://creativecommons.org/licenses/by-sa/3.0/ for more details of
169 Side note: I did text license research some time ago and I think
170 this license is best suited for the purpose at hand and creates the
174 * 1.6 Where is the project website?
176 There is the project website at https://gitlab.com/cryptsetup/cryptsetup/
177 Please do not post questions there, nobody will read them. Use
178 the mailing-list instead.
181 * 1.7 Is there a mailing-list?
183 Instructions on how to subscribe to the mailing-list are at on the
184 project website. People are generally helpful and friendly on the
187 The question of how to unsubscribe from the list does crop up
188 sometimes. For this you need your list management URL, which is
189 sent to you initially and once at the start of each month. Go to
190 the URL mentioned in the email and select "unsubscribe". This page
191 also allows you to request a password reminder.
193 Alternatively, you can send an Email to dm-crypt-request@saout.de
194 with just the word "help" in the subject or message body. Make sure
195 to send it from your list address.
197 The mailing list archive is here:
198 http://dir.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt
201 * 1.8 Unsubscribe from the mailing-list
203 Send mail to dm-crypt-unsubscribe@saout.de from the subscribed
204 account. You will get an email with instructions.
206 Basically, you just have to respond to it unmodified to get
207 unsubscribed. The listserver admin functions are not very fast. It
208 can take 15 minutes or longer for a reply to arrive (I suspect
209 greylisting is in use), so be patient.
211 Also note that nobody on the list can unsubscribe you, sending
212 demands to be unsubscribed to the list just annoys people that are
213 entirely blameless for you being subscribed.
215 If you are subscribed, a subscription confirmation email was sent
216 to your email account and it had to be answered before the
217 subscription went active. The confirmation emails from the
218 listserver have subjects like these (with other numbers):
220 Subject: confirm 9964cf10.....
222 and are sent from dm-crypt-request@saout.de. You should check
223 whether you have anything like it in your sent email folder. If
224 you find nothing and are sure you did not confirm, then you should
225 look into a possible compromise of your email account.
231 * 2.1 LUKS Container Setup mini-HOWTO
233 This item tries to give you a very brief list of all the steps you
234 should go though when creating a new LUKS encrypted container, i.e.
235 encrypted disk, partition or loop-file.
237 01) All data will be lost, if there is data on the target, make a
240 02) Make very sure you have the right target disk, partition or
243 03) If the target was in use previously, it is a good idea to
244 wipe it before creating the LUKS container in order to remove any
245 trace of old file systems and data. For example, some users have
246 managed to run e2fsck on a partition containing a LUKS container,
247 possibly because of residual ext2 superblocks from an earlier use.
248 This can do arbitrary damage up to complete and permanent loss of
249 all data in the LUKS container.
251 To just quickly wipe file systems (old data may remain), use
253 wipefs -a <target device>
255 To wipe file system and data, use something like
257 cat /dev/zero > <target device>
259 This can take a while. To get a progress indicator, you can use
260 the tool dd_rescue (->google) instead or use my stream meter "wcs"
261 (source here: http://www.tansi.org/tools/index.html) in the
264 cat /dev/zero | wcs > <target device>
266 Be very sure you have the right target, all data will be lost!
268 Note that automatic wiping is on the TODO list for cryptsetup, so
269 at some time in the future this will become unnecessary.
271 Alternatively, plain cm-crypt can be used for a very fast wipe with
272 crypto-grade randomness, see Item 2.19
274 04) Create the LUKS container:
275 cryptsetup luksFormat <target device>
277 Just follow the on-screen instructions.
279 Note: Passphrase iteration is determined by cryptsetup depending on
280 CPU power. On a slow device, this may be lower than you want. I
281 recently benchmarked this on a Raspberry Pi and it came out at
282 about 1/15 of the iteration count for a typical PC. If security is
283 paramount, you may want to increase the time spent in iteration, at
284 the cost of a slower unlock later. For the Raspberry Pi, using
286 cryptsetup luksFormat -i 15000 <target device>
288 gives you an iteration count and security level equal to an average
289 PC for passphrase iteration and master-key iteration. If in doubt,
290 check the iteration counts with
292 cryptsetup luksDump <target device>
294 and adjust the iteration count accordingly by creating the container
295 again with a different iteration time (the number after '-i' is the
296 iteration time in milicesonds) until your requirements are met.
298 05) Map the container. Here it will be mapped to /dev/mapper/c1:
299 cryptsetup luksOpen <target device> c1
301 06) (Optionally) wipe the container (make sure you have the right target!):
302 cat /dev/zero > /dev/mapper/c1
304 Note that this creates a small information leak, as an attacker can
305 determine whether a 512 byte block is zero if the attacker has
306 access to the encrypted container multiple times. Typically a
307 competent attacker that has access multiple times can install a
308 passphrase sniffer anyways, so this leakage is not very
309 significant. For getting a progress indicator, see step 03.
311 Note that at some time in the future, cryptsetup will do this for
312 you, but currently it is a TODO list item.
314 07) Create a file system in the mapped container, for example an
315 ext3 file system (any other file system is possible):
317 mke2fs -j /dev/mapper/c1
319 08) Mount your encrypted file system, here on /mnt:
320 mount /dev/mapper/c1 /mnt
322 Done. You can now use the encrypted file system to store data. Be
323 sure to read though the rest of the FAQ, these are just the very
324 basics. In particular, there are a number of mistakes that are
325 easy to make, but will compromise your security.
328 * 2.2 LUKS on partitions or raw disks?
330 This is a complicated question, and made more so by the availability
331 of RAID and LVM. I will try to give some scenarios and discuss
332 advantages and disadvantages. Note that I say LUKS for simplicity,
333 but you can do all the things described with plain dm-crypt as well.
334 Also note that your specific scenario may be so special that most
335 or even all things I say below do not apply.
337 Be aware that if you add LVM into the mix, things can get very
338 complicated. Same with RAID but less so. In particular, data
339 recovery can get exceedingly difficult. Only do so if you have a
340 really good reason and always remember KISS is what separates an
341 engineer from an amateur. Of course, if you really need the added
342 complexity, KISS is satisfied. But be very sure as there is a price
343 to pay for it. In engineering, complexity is always the enemy and
344 needs to be fought without mercy when encountered.
346 Also consider using RAID instead of LVM, as at least with the old
347 superblock format 0.90, the RAID superblock is in the place (end
348 of disk) where the risk of it permanently damaging the LUKS header
349 is smallest and you can have your array assembled by the RAID
350 controller (i.e. the kernel), as it should be. Use partition type
351 0xfd for that. I recommend staying away from superblock formats
352 1.0, 1.1 and 1.2 unless you really need them. Be aware that you
353 lose autodetection with them and have to fall back to some
354 user-space script to do it.
358 (1) Encrypted partition: Just make a partition to your liking,
359 and put LUKS on top of it and a filesystem into the LUKS container.
360 This gives you isolation of differently-tasked data areas, just as
361 ordinary partitioning does. You can have confidential data,
362 non-confidential data, data for some specific applications,
363 user-homes, root, etc. Advantages are simplicity as there is a 1:1
364 mapping between partitions and filesystems, clear security
365 functionality and the ability to separate data into different,
366 independent (!) containers.
368 Note that you cannot do this for encrypted root, that requires an
369 initrd. On the other hand, an initrd is about as vulnerable to a
370 competent attacker as a non-encrypted root, so there really is no
371 security advantage to doing it that way. An attacker that wants to
372 compromise your system will just compromise the initrd or the
373 kernel itself. The better way to deal with this is to make sure the
374 root partition does not store any critical data and move that to
375 additional encrypted partitions. If you really are concerned your
376 root partition may be sabotaged by somebody with physical access
377 (that would however strangely not, say, sabotage your BIOS,
378 keyboard, etc.), protect it in some other way. The PC is just not
379 set-up for a really secure boot-chain (whatever some people may
382 (2) Fully encrypted raw block device: For this, put LUKS on the
383 raw device (e.g. /dev/sdb) and put a filesystem into the LUKS
384 container, no partitioning whatsoever involved. This is very
385 suitable for things like external USB disks used for backups or
386 offline data-storage.
388 (3) Encrypted RAID: Create your RAID from partitions and/or full
389 devices. Put LUKS on top of the RAID device, just if it were an
390 ordinary block device. Applications are just the same as above, but
391 you get redundancy. (Side note as many people seem to be unaware of
392 it: You can do RAID1 with an arbitrary number of components in
393 Linux.) See also Item 2.8.
395 (4) Now, some people advocate doing the encryption below the RAID
396 layer. That has several serious problems. One is that suddenly
397 debugging RAID issues becomes much harder. You cannot do automatic
398 RAID assembly anymore. You need to keep the encryption keys for the
399 components in sync or manage them somehow. The only possible
400 advantage is that things may run a little faster as more CPUs do
401 the encryption, but if speed is a priority over security and
402 simplicity, you are doing this wrong anyways. A good way to
403 mitigate a speed issue is to get a CPU that does hardware AES.
406 * 2.3 How do I set up encrypted swap?
408 As things that are confidential can end up in swap (keys,
409 passphrases, etc. are usually protected against being swapped to
410 disk, but other things may not be), it may be advisable to do
411 something about the issue. One option is to run without swap, which
412 generally works well in a desktop-context. It may cause problems
413 in a server-setting or under special circumstances. The solution to
414 that is to encrypt swap with a random key at boot-time.
416 NOTE: This is for Debian, and should work for Debian-derived
417 distributions. For others you may have to write your own startup
418 script or use other mechanisms.
420 01) Add the swap partition to /etc/crypttab. A line like the following
423 swap /dev/<partition> /dev/urandom swap,noearly
425 Warning: While Debian refuses to overwrite partitions with a
426 filesystem or RAID signature on it, if your disk IDs may change
427 (adding or removing disks, failure of disk during boot, etc.), you
428 may want to take additional precautions. Yes, this means that your
429 kernel device names like sda, sdb, ... can change between reboots!
430 This is not a concern if you have only one disk. One possibility is
431 to make sure the partition number is not present on additional
432 disks or also swap there. Another is to encapsulate the swap
433 partition (by making it a 1-disk RAID1 or by using LVM), so that it
434 gets a persistent identifier. Specifying it directly by UUID does
435 not work, unfortunately, as the UUID is part of the swap signature
436 and that is not visible from the outside due to the encryption and
437 in addition changes on each reboot with this setup.
439 Note: Use /dev/random if you are paranoid or in a potential
440 low-entropy situation (embedded system, etc.). This may cause the
441 operation to take a long time during boot. If you are in a "no
442 entropy" situation, you cannot encrypt swap securely. In this
443 situation you should find some entropy, also because nothing else
444 using crypto will be secure, like ssh, ssl or GnuPG.
446 Note: The "noearly" option makes sure things like LVM, RAID, etc.
447 are running. As swap is non-critical for boot, it is fine to start
450 02) Add the swap partition to /etc/fstab. A line like the following
453 /dev/mapper/swap none swap sw 0 0
455 That is it. Reboot or start it manually to activate encrypted swap.
456 Manual start would look like this:
458 /etc/init.d/crypdisks start
459 swapon /dev/mapper/swap
462 * 2.4 What is the difference between "plain" and LUKS format?
464 First, unless you happen to understand the cryptographic background
465 well, you should use LUKS. It does protect the user from a lot of
466 common mistakes. Plain dm-crypt is for experts.
468 Plain format is just that: It has no metadata on disk, reads all
469 parameters from the commandline (or the defaults), derives a
470 master-key from the passphrase and then uses that to de-/encrypt
471 the sectors of the device, with a direct 1:1 mapping between
472 encrypted and decrypted sectors.
474 Primary advantage is high resilience to damage, as one damaged
475 encrypted sector results in exactly one damaged decrypted sector.
476 Also, it is not readily apparent that there even is encrypted data
477 on the device, as an overwrite with crypto-grade randomness (e.g.
478 from /dev/urandom) looks exactly the same on disk.
480 Side-note: That has limited value against the authorities. In
481 civilized countries, they cannot force you to give up a crypto-key
482 anyways. In quite a few countries around the world, they can force
483 you to give up the keys (using imprisonment or worse to pressure
484 you, sometimes without due process), and in the worst case, they
485 only need a nebulous "suspicion" about the presence of encrypted
486 data. Sometimes this applies to everybody, sometimes only when you
487 are suspected of having "illicit data" (definition subject to
488 change) and sometimes specifically when crossing a border. Note
489 that this is going on in countries like the US and the UK, to
490 different degrees and sometimes with courts restricting what the
491 authorities can actually demand.
493 My advice is to either be ready to give up the keys or to not have
494 encrypted data when traveling to those countries, especially when
495 crossing the borders. The latter also means not having any
496 high-entropy (random) data areas on your disk, unless you can
497 explain them and demonstrate that explanation. Hence doing a
498 zero-wipe of all free space, including unused space, may be a good
501 Disadvantages are that you do not have all the nice features that
502 the LUKS metadata offers, like multiple passphrases that can be
503 changed, the cipher being stored in the metadata, anti-forensic
504 properties like key-slot diffusion and salts, etc..
506 LUKS format uses a metadata header and 8 key-slot areas that are
507 being placed at the beginning of the disk, see below under "What
508 does the LUKS on-disk format looks like?". The passphrases are used
509 to decrypt a single master key that is stored in the anti-forensic
512 Advantages are a higher usability, automatic configuration of
513 non-default crypto parameters, defenses against low-entropy
514 passphrases like salting and iterated PBKDF2 passphrase hashing,
515 the ability to change passphrases, and others.
517 Disadvantages are that it is readily obvious there is encrypted
518 data on disk (but see side note above) and that damage to the
519 header or key-slots usually results in permanent data-loss. See
520 below under "6. Backup and Data Recovery" on how to reduce that
521 risk. Also the sector numbers get shifted by the length of the
522 header and key-slots and there is a loss of that size in capacity
523 (1MB+4096B for defaults and 2MB for the most commonly used
524 non-default XTS mode).
527 * 2.5 Can I encrypt an already existing, non-empty partition to use
530 There is no converter, and it is not really needed. The way to do
531 this is to make a backup of the device in question, securely wipe
532 the device (as LUKS device initialization does not clear away old
533 data), do a luksFormat, optionally overwrite the encrypted device,
534 create a new filesystem and restore your backup on the now
535 encrypted device. Also refer to sections "Security Aspects" and
536 "Backup and Data Recovery".
538 For backup, plain GNU tar works well and backs up anything likely
539 to be in a filesystem.
542 * 2.6 How do I use LUKS with a loop-device?
544 This can be very handy for experiments. Setup is just the same as
545 with any block device. If you want, for example, to use a 100MiB
546 file as LUKS container, do something like this:
548 head -c 100M /dev/zero > luksfile # create empty file
549 losetup /dev/loop0 luksfile # map luksfile to /dev/loop0
550 cryptsetup luksFormat /dev/loop0 # create LUKS on loop device
552 Afterwards just use /dev/loop0 as a you would use a LUKS partition.
553 To unmap the file when done, use "losetup -d /dev/loop0".
556 * 2.7 When I add a new key-slot to LUKS, it asks for a passphrase but
557 then complains about there not being a key-slot with that
560 That is as intended. You are asked a passphrase of an existing
561 key-slot first, before you can enter the passphrase for the new
562 key-slot. Otherwise you could break the encryption by just adding a
563 new key-slot. This way, you have to know the passphrase of one of
564 the already configured key-slots in order to be able to configure a
568 * 2.8 Encryption on top of RAID or the other way round?
570 Unless you have special needs, place encryption between RAID and
571 filesystem, i.e. encryption on top of RAID. You can do it the other
572 way round, but you have to be aware that you then need to give the
573 passphrase for each individual disk and RAID autodetection will
574 not work anymore. Therefore it is better to encrypt the RAID
575 device, e.g. /dev/dm0 .
577 This means that the typical layering looks like this:
589 The big advantage is that you can manage the RAID container just
590 like any RAID container, it does not care that what is in it is
594 * 2.9 How do I read a dm-crypt key from file?
596 Use the --key-file option, like this:
598 cryptsetup create --key-file keyfile e1 /dev/loop0
600 This will read the binary key from file, i.e. no hashing or
601 transformation will be applied to the keyfile before its bits are
602 used as key. Extra bits (beyond the length of the key) at the end
603 are ignored. Note that if you read from STDIN, the data will still
604 be hashed, just as a key read interactively from the terminal. See
605 the man-page sections "NOTES ON PASSPHRASE PROCESSING..." for more
609 * 2.10 How do I read a LUKS slot key from file?
611 What you really do here is to read a passphrase from file, just as
612 you would with manual entry of a passphrase for a key-slot. You can
613 add a new passphrase to a free key-slot, set the passphrase of an
614 specific key-slot or put an already configured passphrase into a
615 file. In the last case make sure no trailing newline (0x0a) is
616 contained in the key file, or the passphrase will not work because
617 the whole file is used as input.
619 To add a new passphrase to a free key slot from file, use something
622 cryptsetup luksAddKey /dev/loop0 keyfile
624 To add a new passphrase to a specific key-slot, use something like
627 cryptsetup luksAddKey --key-slot 7 /dev/loop0 keyfile
629 To supply a key from file to any LUKS command, use the --key-file
630 option, e.g. like this:
632 cryptsetup luksOpen --key-file keyfile /dev/loop0 e1
635 * 2.11 How do I read the LUKS master key from file?
637 The question you should ask yourself first is why you would want to
638 do this. The only legitimate reason I can think of is if you want
639 to have two LUKS devices with the same master key. Even then, I
640 think it would be preferable to just use key-slots with the same
641 passphrase, or to use plain dm-crypt instead. If you really have a
642 good reason, please tell me. If I am convinced, I will add how to
646 * 2.12 What are the security requirements for a key read from file?
648 A file-stored key or passphrase has the same security requirements
649 as one entered interactively, however you can use random bytes and
650 thereby use bytes you cannot type on the keyboard. You can use any
651 file you like as key file, for example a plain text file with a
652 human readable passphrase. To generate a file with random bytes,
653 use something like this:
655 head -c 256 /dev/random > keyfile
658 * 2.13 If I map a journaled file system using dm-crypt/LUKS, does it
659 still provide its usual transactional guarantees?
661 Yes, it does, unless a very old kernel is used. The required flags
662 come from the filesystem layer and are processed and passed onwards
663 by dm-crypt. A bit more information on the process by which
664 transactional guarantees are implemented can be found here:
666 http://lwn.net/Articles/400541/
668 Please note that these "guarantees" are weaker than they appear to
669 be. One problem is that quite a few disks lie to the OS about
670 having flushed their buffers. Some other things can go wrong as
671 well. The filesystem developers are aware of these problems and
672 typically can make it work anyways. That said, dm-crypt/LUKS will
673 not make things worse.
675 One specific problem you can run into though is that you can get
676 short freezes and other slowdowns due to the encryption layer.
677 Encryption takes time and forced flushes will block for that time.
678 For example, I did run into frequent small freezes (1-2 sec) when
679 putting a vmware image on ext3 over dm-crypt. When I went back to
680 ext2, the problem went away. This seems to have gotten better with
681 kernel 2.6.36 and the reworking of filesystem flush locking
682 mechanism (less blocking of CPU activity during flushes). It
683 should improve further and eventually the problem should go away.
686 * 2.14 Can I use LUKS or cryptsetup with a more secure (external)
687 medium for key storage, e.g. TPM or a smartcard?
689 Yes, see the answers on using a file-supplied key. You do have to
690 write the glue-logic yourself though. Basically you can have
691 cryptsetup read the key from STDIN and write it there with your
692 own tool that in turn gets the key from the more secure key
695 For TPM support, you may want to have a look at tpm-luks at
696 https://github.com/shpedoikal/tpm-luks. Note that tpm-luks is not
697 related to the cryptsetup project.
700 * 2.15 Can I resize a dm-crypt or LUKS partition?
702 Yes, you can, as neither dm-crypt nor LUKS stores partition size.
703 Whether you should is a different question. Personally I recommend
704 backup, recreation of the encrypted partition with new size,
705 recreation of the filesystem and restore. This gets around the
706 tricky business of resizing the filesystem. Resizing a dm-crypt or
707 LUKS container does not resize the filesystem in it. The backup is
708 really non-optional here, as a lot can go wrong, resulting in
709 partial or complete data loss. Using something like gparted to
710 resize an encrypted partition is slow, but typically works. This
711 will not change the size of the filesystem hidden under the
714 You also need to be aware of size-based limitations. The one
715 currently relevant is that aes-xts-plain should not be used for
716 encrypted container sizes larger than 2TiB. Use aes-xts-plain64
720 * 2.16 How do I Benchmark the Ciphers, Hashes and Modes?
722 Since version 1.60 cryptsetup supports the "benchmark" command.
727 It will output first iterations/second for the key-derivation
728 function PBKDF2 parameterized with different hash-functions, and
729 then the raw encryption speed of ciphers with different modes and
730 key-sizes. You can get more than the default benchmarks, see the
731 man-page for the relevant parameters. Note that XTS mode takes two
732 keys, hence the listed key sizes are double that for other modes
733 and half of it is the cipher key, the other half is the XTS key.
736 * 2.17 How do I Verify I have an Authentic cryptsetup Source Package?
738 Current maintainer is Milan Broz and he signs the release packages
739 with his PGP key. The key he currently uses is the "RSA key ID
740 D93E98FC", fingerprint 2A29 1824 3FDE 4664 8D06 86F9 D9B0 577B
741 D93E 98FC. While I have every confidence this really is his key and
742 that he is who he claims to be, don't depend on it if your life is
743 at stake. For that matter, if your life is at stake, don't depend
744 on me being who I claim to be either.
746 That said, as cryptsetup is under good version control, a malicious
747 change should be noticed sooner or later, but it may take a while.
748 Also, the attacker model makes compromising the sources in a
749 non-obvious way pretty hard. Sure, you could put the master-key
750 somewhere on disk, but that is rather obvious as soon as somebody
751 looks as there would be data in an empty LUKS container in a place
752 it should not be. Doing this in a more nefarious way, for example
753 hiding the master-key in the salts, would need a look at the
754 sources to be discovered, but I think that somebody would find that
755 sooner or later as well.
757 That said, this discussion is really a lot more complicated and
758 longer as an FAQ can sustain. If in doubt, ask on the mailing list.
761 * 2.18 Is there a concern with 4k Sectors?
763 Not from dm-crypt itself. Encryption will be done in 512B blocks,
764 but if the partition and filesystem are aligned correctly and the
765 filesystem uses multiples of 4kiB as block size, the dm-crypt layer
766 will just process 8 x 512B = 4096B at a time with negligible
767 overhead. LUKS does place data at an offset, which is 2MiB per
768 default and will not break alignment. See also Item 6.12 of this
769 FAQ for more details. Note that if your partition or filesystem is
770 misaligned, dm-crypt can make the effect worse though.
773 * 2.19 How can I wipe a device with crypto-grade randomness?
775 The conventional recommendation if you want to not just do a
776 zero-wipe is to use something like
778 cat /dev/urandom > <taget-device>
780 That is very slow and painful at 10-20MB/s on a fast computer.
781 Using cryptsetup and a plain dm-crypt device with a random key, it
782 is much faster and gives you the same level of security. The
783 defaults are quite enough.
785 For device set-up, do the following:
787 cryptsetup open --type plain -d /dev/urandom /dev/<block-device> to_be_wiped
789 Then you have several options. Simple wipe without
792 cat /dev/zero > /dev/mapper/to_be_wiped
794 Progress-indicator by dd_rescue:
796 dd_rescue -w /dev/zero /dev/mapper/to_be_wiped
798 Progress-indicator by my "wcs" stream meter (available from
799 http://www.tansi.org/tools/index.html ):
801 cat /dev/zero | wcs > /dev/mapper/to_be_wiped
803 Remove the mapping at the end and you are done.
809 * 3.1 My dm-crypt/LUKS mapping does not work! What general steps are
810 there to investigate the problem?
812 If you get a specific error message, investigate what it claims
813 first. If not, you may want to check the following things.
815 - Check that "/dev", including "/dev/mapper/control" is there. If it
816 is missing, you may have a problem with the "/dev" tree itself or
817 you may have broken udev rules.
819 - Check that you have the device mapper and the crypt target in your
820 kernel. The output of "dmsetup targets" should list a "crypt"
821 target. If it is not there or the command fails, add device mapper
822 and crypt-target to the kernel.
824 - Check that the hash-functions and ciphers you want to use are in
825 the kernel. The output of "cat /proc/crypto" needs to list them.
828 * 3.2 My dm-crypt mapping suddenly stopped when upgrading cryptsetup.
830 The default cipher, hash or mode may have changed (the mode changed
831 from 1.0.x to 1.1.x). See under "Issues With Specific Versions of
835 * 3.3 When I call cryptsetup from cron/CGI, I get errors about
838 If you get errors about unknown parameters or the like that are not
839 present when cryptsetup is called from the shell, make sure you
840 have no older version of cryptsetup on your system that then gets
841 called by cron/CGI. For example some distributions install
842 cryptsetup into /usr/sbin, while a manual install could go to
843 /usr/local/sbin. As a debugging aid, call "cryptsetup --version"
844 from cron/CGI or the non-shell mechanism to be sure the right
848 * 3.4 Unlocking a LUKS device takes very long. Why?
850 The iteration time for a key-slot (see Section 5 for an explanation
851 what iteration does) is calculated when setting a passphrase. By
852 default it is 1 second on the machine where the passphrase is set.
853 If you set a passphrase on a fast machine and then unlock it on a
854 slow machine, the unlocking time can be much longer. Also take into
855 account that up to 8 key-slots have to be tried in order to find the
858 If this is problem, you can add another key-slot using the slow
859 machine with the same passphrase and then remove the old key-slot.
860 The new key-slot will have an iteration count adjusted to 1 second
861 on the slow machine. Use luksKeyAdd and then luksKillSlot or
864 However, this operation will not change volume key iteration count
865 (MK iterations in output of "cryptsetup luksDump"). In order to
866 change that, you will have to backup the data in the LUKS
867 container (i.e. your encrypted data), luksFormat on the slow
868 machine and restore the data. Note that in the original LUKS
869 specification this value was fixed to 10, but it is now derived
870 from the PBKDF2 benchmark as well and set to iterations in 0.125
871 sec or 1000, whichever is larger. Also note that MK iterations
872 are not very security relevant. But as each key-slot already takes
873 1 second, spending the additional 0.125 seconds really does not
877 * 3.5 "blkid" sees a LUKS UUID and an ext2/swap UUID on the same
878 device. What is wrong?
880 Some old versions of cryptsetup have a bug where the header does
881 not get completely wiped during LUKS format and an older ext2/swap
882 signature remains on the device. This confuses blkid.
884 Fix: Wipe the unused header areas by doing a backup and restore of
885 the header with cryptsetup 1.1.x:
887 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
888 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
891 * 3.6 cryptsetup segfaults on Gentoo amd64 hardened ...
893 There seems to be some interference between the hardening and and
894 the way cryptsetup benchmarks PBKDF2. The solution to this is
895 currently not quite clear for an encrypted root filesystem. For
896 other uses, you can apparently specify USE="dynamic" as compile
897 flag, see http://bugs.gentoo.org/show_bug.cgi?id=283470
903 * 4.1 I get the error "LUKS keyslot x is invalid." What does that
906 This means that the given keyslot has an offset that points
907 outside the valid keyslot area. Typically, the reason is a
908 corrupted LUKS header because something was written to the start of
909 the device the LUKS container is on. Refer to Section "Backup and
910 Data Recovery" and ask on the mailing list if you have trouble
911 diagnosing and (if still possible) repairing this.
914 * 4.2 I cannot unlock my LUKS container! What could be the problem?
916 First, make sure you have a correct passphrase. Then make sure you
917 have the correct key-map and correct keyboard. And then make sure
918 you have the correct character set and encoding, see also
919 "PASSPHRASE CHARACTER SET" under Section 1.2.
921 If you are sure you are entering the passphrase right, there is the
922 possibility that the respective key-slot has been damaged. There
923 is no way to recover a damaged key-slot, except from a header
924 backup (see Section 6). For security reasons, there is also no
925 checksum in the key-slots that could tell you whether a key-slot has
926 been damaged. The only checksum present allows recognition of a
927 correct passphrase, but that only works if the passphrase is
928 correct and the respective key-slot is intact.
930 In order to find out whether a key-slot is damaged one has to look
931 for "non-random looking" data in it. There is a tool that
932 automatizes this in the cryptsetup distribution from version 1.6.0
933 onwards. It is located in misc/keyslot_checker/. Instructions how
934 to use and how to interpret results are in the README file. Note
935 that this tool requires a libcryptsetup from cryptsetup 1.6.0 or
936 later (which means libcryptsetup.so.4.5.0 or later). If the tool
937 complains about missing functions in libcryptsetup, you likely
938 have an earlier version from your distribution still installed. You
939 can either point the symbolic link(s) from libcryptsetup.so.4 to
940 the new version manually, or you can uninstall the distribution
941 version of cryptsetup and re-install that from cryptsetup >= 1.6.0
945 * 4.3 Can a bad RAM module cause problems?
947 LUKS and dm-crypt can give the RAM quite a workout, especially when
948 combined with software RAID. In particular the combination RAID5 +
949 LUKS + XFS seems to uncover RAM problems that never caused obvious
950 problems before. Symptoms vary, but often the problem manifest
951 itself when copying large amounts of data, typically several times
952 larger than your main memory.
954 Side note: One thing you should always do on large data
955 copy/movements is to run a verify, for example with the "-d"
956 option of "tar" or by doing a set of MD5 checksums on the source
959 find . -type f -exec md5sum \{\} \; > checksum-file
961 and then a "md5sum -c checksum-file" on the other side. If you get
962 mismatches here, RAM is the primary suspect. A lesser suspect is
963 an overclocked CPU. I have found countless hardware problems in
964 verify runs after copying or making backups. Bit errors are much
965 more common than most people think.
967 Some RAM issues are even worse and corrupt structures in one of the
968 layers. This typically results in lockups, CPU state dumps in the
969 system logs, kernel panic or other things. It is quite possible to
970 have the problem with an encrypted device, but not with an
971 otherwise the same unencrypted device. The reason for that is that
972 encryption has an error amplification property: You flip one bit
973 in an encrypted data block, and the decrypted version has half of
974 its bits flipped. This is an important security property for modern
975 ciphers. With the usual modes in cryptsetup (CBC, ESSIV, XTS), you
976 get up to a completely changed 512 byte block per bit error. A
977 corrupt block causes a lot more havoc than the occasionally
978 flipped single bit and can result in various obscure errors.
980 Note, that a verify run on copying between encrypted or
981 unencrypted devices will reliably detect corruption, even when the
982 copying itself did not report any problems. If you find defect
983 RAM, assume all backups and copied data to be suspect, unless you
987 * 4.4 How do I test RAM?
989 First you should know that overclocking often makes memory
990 problems worse. So if you overclock (which I strongly recommend
991 against in a system holding data that has some worth), run the
992 tests with the overclocking active.
994 There are two good options. One is Memtest86+ and the other is
995 "memtester" by Charles Cazabon. Memtest86+ requires a reboot and
996 then takes over the machine, while memtester runs from a
997 root-shell. Both use different testing methods and I have found
998 problems fast with each one that the other needed long to find. I
999 recommend running the following procedure until the first error is
1002 - Run Memtest86+ for one cycle
1004 - Run memtester for one cycle (shut down as many other applications
1007 - Run Memtest86+ for 24h or more
1009 - Run memtester for 24h or more
1011 If all that does not produce error messages, your RAM may be sound,
1012 but I have had one weak bit that Memtest86+ needed around 60 hours
1013 to find. If you can reproduce the original problem reliably, a good
1014 additional test may be to remove half of the RAM (if you have more
1015 than one module) and try whether the problem is still there and if
1016 so, try with the other half. If you just have one module, get a
1017 different one and try with that. If you do overclocking, reduce
1018 the settings to the most conservative ones available and try with
1025 * 5.1 How long is a secure passphrase ?
1027 This is just the short answer. For more info and explanation of
1028 some of the terms used in this item, read the rest of Section 5.
1029 The actual recommendation is at the end of this item.
1031 First, passphrase length is not really the right measure,
1032 passphrase entropy is. For example, a random lowercase letter (a-z)
1033 gives you 4.7 bit of entropy, one element of a-z0-9 gives you 5.2
1034 bits of entropy, an element of a-zA-Z0-9 gives you 5.9 bits and
1035 a-zA-Z0-9!@#$%^&:-+ gives you 6.2 bits. On the other hand, a random
1036 English word only gives you 0.6...1.3 bits of entropy per
1037 character. Using sentences that make sense gives lower entropy,
1038 series of random words gives higher entropy. Do not use sentences
1039 that can be tied to you or found on your computer. This type of
1040 attack is done routinely today.
1042 That said, it does not matter too much what scheme you use, but it
1043 does matter how much entropy your passphrase contains, because an
1044 attacker has to try on average
1046 1/2 * 2^(bits of entropy in passphrase)
1048 different passphrases to guess correctly.
1050 Historically, estimations tended to use computing time estimates,
1051 but more modern approaches try to estimate cost of guessing a
1054 As an example, I will try to get an estimate from the numbers in
1055 http://it.slashdot.org/story/12/12/05/0623215/new-25-gpu-monster-devours-strong-passwords-in-minutes
1056 More references can be found a the end of this document. Note that
1057 these are estimates from the defender side, so assuming something
1058 is easier than it actually is is fine. An attacker may still have
1059 vastly higher cost than estimated here.
1061 LUKS uses SHA1 for hashing per default. The claim in the reference
1062 is 63 billion tries/second for SHA1. We will leave aside the check
1063 whether a try actually decrypts a key-slot. Now, the machine has 25
1064 GPUs, which I will estimate at an overall lifetime cost of USD/EUR
1065 1000 each, and an useful lifetime of 2 years. (This is on the low
1066 side.) Disregarding downtime, the machine can then break
1068 N = 63*10^9 * 3600 * 24 * 365 * 2 ~ 4*10^18
1070 passphrases for EUR/USD 25k. That is one 62 bit passphrase hashed
1071 once with SHA1 for EUR/USD 25k. Note that as this can be
1072 parallelized, it can be done faster than 2 years with several of
1075 For plain dm-crypt (no hash iteration) this is it. This gives (with
1076 SHA1, plain dm-crypt default is ripemd160 which seems to be
1077 slightly slower than SHA1):
1079 Passphrase entropy Cost to break
1088 For LUKS, you have to take into account hash iteration in PBKDF2.
1089 For a current CPU, there are about 100k iterations (as can be
1090 queried with ''cryptsetup luksDump''.
1092 The table above then becomes:
1094 Passphrase entropy Cost to break
1105 To get reasonable security for the next 10 years, it is a good idea
1106 to overestimate by a factor of at least 1000.
1108 Then there is the question of how much the attacker is willing to
1109 spend. That is up to your own security evaluation. For general use,
1110 I will assume the attacker is willing to spend up to 1 million
1111 EUR/USD. Then we get the following recommendations:
1113 Plain dm-crypt: Use > 80 bit. That is e.g. 17 random chars from a-z
1114 or a random English sentence of > 135 characters length.
1116 LUKS: Use > 65 bit. That is e.g. 14 random chars from a-z or a
1117 random English sentence of > 108 characters length.
1119 If paranoid, add at least 20 bit. That is roughly four additional
1120 characters for random passphrases and roughly 32 characters for a
1121 random English sentence.
1124 * 5.2 Is LUKS insecure? Everybody can see I have encrypted data!
1126 In practice it does not really matter. In most civilized countries
1127 you can just refuse to hand over the keys, no harm done. In some
1128 countries they can force you to hand over the keys, if they suspect
1129 encryption. However the suspicion is enough, they do not have to
1130 prove anything. This is for practical reasons, as even the presence
1131 of a header (like the LUKS header) is not enough to prove that you
1132 have any keys. It might have been an experiment, for example. Or it
1133 was used as encrypted swap with a key from /dev/random. So they
1134 make you prove you do not have encrypted data. Of course that is
1135 just as impossible as the other way round.
1137 This means that if you have a large set of random-looking data,
1138 they can already lock you up. Hidden containers (encryption hidden
1139 within encryption), as possible with Truecrypt, do not help
1140 either. They will just assume the hidden container is there and
1141 unless you hand over the key, you will stay locked up. Don't have
1142 a hidden container? Though luck. Anybody could claim that.
1144 Still, if you are concerned about the LUKS header, use plain
1145 dm-crypt with a good passphrase. See also Section 2, "What is the
1146 difference between "plain" and LUKS format?"
1149 * 5.3 Should I initialize (overwrite) a new LUKS/dm-crypt partition?
1151 If you just create a filesystem on it, most of the old data will
1152 still be there. If the old data is sensitive, you should overwrite
1153 it before encrypting. In any case, not initializing will leave the
1154 old data there until the specific sector gets written. That may
1155 enable an attacker to determine how much and where on the
1156 partition data was written. If you think this is a risk, you can
1157 prevent this by overwriting the encrypted device (here assumed to
1158 be named "e1") with zeros like this:
1160 dd_rescue -w /dev/zero /dev/mapper/e1
1162 or alternatively with one of the following more standard commands:
1164 cat /dev/zero > /dev/mapper/e1
1165 dd if=/dev/zero of=/dev/mapper/e1
1168 * 5.4 How do I securely erase a LUKS (or other) partition?
1170 For LUKS, if you are in a desperate hurry, overwrite the LUKS
1171 header and key-slot area. This means overwriting the first
1172 (keyslots x stripes x keysize) + offset bytes. For the default
1173 parameters, this is the 1'052'672 bytes, i.e. 1MiB + 4096 of the
1174 LUKS partition. For 512 bit key length (e.g. for aes-xts-plain with
1175 512 bit key) this is 2MiB. (The different offset stems from
1176 differences in the sector alignment of the key-slots.) If in doubt,
1177 just be generous and overwrite the first 10MB or so, it will likely
1178 still be fast enough. A single overwrite with zeros should be
1179 enough. If you anticipate being in a desperate hurry, prepare the
1180 command beforehand. Example with /dev/sde1 as the LUKS partition
1181 and default parameters:
1183 head -c 1052672 /dev/zero > /dev/sde1; sync
1185 A LUKS header backup or full backup will still grant access to
1186 most or all data, so make sure that an attacker does not have
1187 access to backups or destroy them as well.
1189 If you have time, overwrite the whole LUKS partition with a single
1190 pass of zeros. This is enough for current HDDs. For SSDs or FLASH
1191 (USB sticks) you may want to overwrite the whole drive several
1192 times to be sure data is not retained by wear leveling. This is
1193 possibly still insecure as SSD technology is not fully understood
1194 in this regard. Still, due to the anti-forensic properties of the
1195 LUKS key-slots, a single overwrite of an SSD or FLASH drive could
1196 be enough. If in doubt, use physical destruction in addition. Here
1197 is a link to some current research results on erasing SSDs and
1199 http://www.usenix.org/events/fast11/tech/full_papers/Wei.pdf
1201 Keep in mind to also erase all backups.
1203 Example for a zero-overwrite erase of partition sde1 done with
1206 dd_rescue -w /dev/zero /dev/sde1
1209 * 5.5 How do I securely erase a backup of a LUKS partition or header?
1211 That depends on the medium it is stored on. For HDD and SSD, use
1212 overwrite with zeros. For an SSD or FLASH drive (USB stick), you
1213 may want to overwrite the complete SSD several times and use
1214 physical destruction in addition, see last item. For re-writable
1215 CD/DVD, a single overwrite should also be enough, due to the
1216 anti-forensic properties of the LUKS keyslots. For write-once
1217 media, use physical destruction. For low security requirements,
1218 just cut the CD/DVD into several parts. For high security needs,
1219 shred or burn the medium. If your backup is on magnetic tape, I
1220 advise physical destruction by shredding or burning, after
1221 overwriting . The problem with magnetic tape is that it has a
1222 higher dynamic range than HDDs and older data may well be
1223 recoverable after overwrites. Also write-head alignment issues can
1224 lead to data not actually being deleted at all during overwrites.
1227 * 5.6 What about backup? Does it compromise security?
1229 That depends. See item 6.7.
1232 * 5.7 Why is all my data permanently gone if I overwrite the LUKS
1235 Overwriting the LUKS header in part or in full is the most common
1236 reason why access to LUKS containers is lost permanently.
1237 Overwriting can be done in a number of fashions, like creating a
1238 new filesystem on the raw LUKS partition, making the raw partition
1239 part of a raid array and just writing to the raw partition.
1241 The LUKS header contains a 256 bit "salt" per key-slot and without
1242 that no decryption is possible. While the salts are not secret,
1243 they are key-grade material and cannot be reconstructed. This is a
1244 cryptographically strong "cannot". From observations on the
1245 cryptsetup mailing-list, people typically go though the usual
1246 stages of grief (Denial, Anger, Bargaining, Depression, Acceptance)
1247 when this happens to them. Observed times vary between 1 day and 2
1248 weeks to complete the cycle. Seeking help on the mailing-list is
1249 fine. Even if we usually cannot help with getting back your data,
1250 most people found the feedback comforting.
1252 If your header does not contain an intact key-slot salt, best go
1253 directly to the last stage ("Acceptance") and think about what to
1254 do now. There is one exception that I know of: If your LUKS
1255 container is still open, then it may be possible to extract the
1256 master key from the running system. See Item "How do I recover the
1257 master key from a mapped LUKS container?" in Section "Backup and
1261 * 5.8 What is a "salt"?
1263 A salt is a random key-grade value added to the passphrase before
1264 it is processed. It is not kept secret. The reason for using salts
1265 is as follows: If an attacker wants to crack the password for a
1266 single LUKS container, then every possible passphrase has to be
1267 tried. Typically an attacker will not try every binary value, but
1268 will try words and sentences from a dictionary.
1270 If an attacker wants to attack several LUKS containers with the
1271 same dictionary, then a different approach makes sense: Compute the
1272 resulting slot-key for each dictionary element and store it on
1273 disk. Then the test for each entry is just the slow unlocking with
1274 the slot key (say 0.00001 sec) instead of calculating the slot-key
1275 first (1 sec). For a single attack, this does not help. But if you
1276 have more than one container to attack, this helps tremendously,
1277 also because you can prepare your table before you even have the
1278 container to attack! The calculation is also very simple to
1279 parallelize. You could, for example, use the night-time unused CPU
1280 power of your desktop PCs for this.
1282 This is where the salt comes in. If the salt is combined with the
1283 passphrase (in the simplest form, just appended to it), you
1284 suddenly need a separate table for each salt value. With a
1285 reasonably-sized salt value (256 bit, e.g.) this is quite
1289 * 5.9 Is LUKS secure with a low-entropy (bad) passphrase?
1291 Note: You should only use the 94 printable characters from 7 bit
1292 ASCII code to prevent your passphrase from failing when the
1293 character encoding changes, e.g. because of a system upgrade, see
1294 also the note at the very start of this FAQ under "WARNINGS".
1296 This needs a bit of theory. The quality of your passphrase is
1297 directly related to its entropy (information theoretic, not
1298 thermodynamic). The entropy says how many bits of "uncertainty" or
1299 "randomness" are in you passphrase. In other words, that is how
1300 difficult guessing the passphrase is.
1302 Example: A random English sentence has about 1 bit of entropy per
1303 character. A random lowercase (or uppercase) character has about
1306 Now, if n is the number of bits of entropy in your passphrase and t
1307 is the time it takes to process a passphrase in order to open the
1308 LUKS container, then an attacker has to spend at maximum
1310 attack_time_max = 2^n * t
1312 time for a successful attack and on average half that. There is no
1313 way getting around that relationship. However, there is one thing
1314 that does help, namely increasing t, the time it takes to use a
1315 passphrase, see next FAQ item.
1317 Still, if you want good security, a high-entropy passphrase is the
1318 only option. For example, a low-entropy passphrase can never be
1319 considered secure against a TLA-level (Three Letter Agency level,
1320 i.e. government-level) attacker, no matter what tricks are used in
1321 the key-derivation function. Use at least 64 bits for secret stuff.
1322 That is 64 characters of English text (but only if randomly chosen)
1323 or a combination of 12 truly random letters and digits.
1325 For passphrase generation, do not use lines from very well-known
1326 texts (religious texts, Harry potter, etc.) as they are to easy to
1327 guess. For example, the total Harry Potter has about 1'500'000
1328 words (my estimation). Trying every 64 character sequence starting
1329 and ending at a word boundary would take only something like 20
1330 days on a single CPU and is entirely feasible. To put that into
1331 perspective, using a number of Amazon EC2 High-CPU Extra Large
1332 instances (each gives about 8 real cores), this test costs
1333 currently about 50USD/EUR, but can be made to run arbitrarily fast.
1335 On the other hand, choosing 1.5 lines from, say, the Wheel of Time
1336 is in itself not more secure, but the book selection adds quite a
1337 bit of entropy. (Now that I have mentioned it here, don't use tWoT
1338 either!) If you add 2 or 3 typos or switch some words around, then
1339 this is good passphrase material.
1342 * 5.10 What is "iteration count" and why is decreasing it a bad idea?
1344 Iteration count is the number of PBKDF2 iterations a passphrase is
1345 put through before it is used to unlock a key-slot. Iterations are
1346 done with the explicit purpose to increase the time that it takes
1347 to unlock a key-slot. This provides some protection against use of
1348 low-entropy passphrases.
1350 The idea is that an attacker has to try all possible passphrases.
1351 Even if the attacker knows the passphrase is low-entropy (see last
1352 item), it is possible to make each individual try take longer. The
1353 way to do this is to repeatedly hash the passphrase for a certain
1354 time. The attacker then has to spend the same time (given the same
1355 computing power) as the user per try. With LUKS, the default is 1
1356 second of PBKDF2 hashing.
1358 Example 1: Lets assume we have a really bad passphrase (e.g. a
1359 girlfriends name) with 10 bits of entropy. With the same CPU, an
1360 attacker would need to spend around 500 seconds on average to
1361 break that passphrase. Without iteration, it would be more like
1362 0.0001 seconds on a modern CPU.
1364 Example 2: The user did a bit better and has 32 chars of English
1365 text. That would be about 32 bits of entropy. With 1 second
1366 iteration, that means an attacker on the same CPU needs around 136
1367 years. That is pretty impressive for such a weak passphrase.
1368 Without the iterations, it would be more like 50 days on a modern
1369 CPU, and possibly far less.
1371 In addition, the attacker can both parallelize and use special
1372 hardware like GPUs or FPGAs to speed up the attack. The attack can
1373 also happen quite some time after the luksFormat operation and CPUs
1374 can have become faster and cheaper. For that reason you want a
1375 bit of extra security. Anyways, in Example 1 your are screwed.
1376 In example 2, not necessarily. Even if the attack is faster, it
1377 still has a certain cost associated with it, say 10000 EUR/USD
1378 with iteration and 1 EUR/USD without iteration. The first can be
1379 prohibitively expensive, while the second is something you try
1380 even without solid proof that the decryption will yield something
1383 The numbers above are mostly made up, but show the idea. Of course
1384 the best thing is to have a high-entropy passphrase.
1386 Would a 100 sec iteration time be even better? Yes and no.
1387 Cryptographically it would be a lot better, namely 100 times better.
1388 However, usability is a very important factor for security
1389 technology and one that gets overlooked surprisingly often. For
1390 LUKS, if you have to wait 2 minutes to unlock the LUKS container,
1391 most people will not bother and use less secure storage instead. It
1392 is better to have less protection against low-entropy passphrases
1393 and people actually use LUKS, than having them do without
1394 encryption altogether.
1396 Now, what about decreasing the iteration time? This is generally a
1397 very bad idea, unless you know and can enforce that the users only
1398 use high-entropy passphrases. If you decrease the iteration time
1399 without ensuring that, then you put your users at increased risk,
1400 and considering how rarely LUKS containers are unlocked in a
1401 typical work-flow, you do so without a good reason. Don't do it.
1402 The iteration time is already low enough that users with entropy
1403 low passphrases are vulnerable. Lowering it even further increases
1404 this danger significantly.
1407 * 5.11 Some people say PBKDF2 is insecure?
1409 There is some discussion that a hash-function should have a "large
1410 memory" property, i.e. that it should require a lot of memory to be
1411 computed. This serves to prevent attacks using special programmable
1412 circuits, like FPGAs, and attacks using graphics cards. PBKDF2
1413 does not need a lot of memory and is vulnerable to these attacks.
1414 However, the publication usually referred in these discussions is
1415 not very convincing in proving that the presented hash really is
1416 "large memory" (that may change, email the FAQ maintainer when it
1417 does) and it is of limited usefulness anyways. Attackers that use
1418 clusters of normal PCs will not be affected at all by a "large
1419 memory" property. For example the US Secret Service is known to
1420 use the off-hour time of all the office PCs of the Treasury for
1421 password breaking. The Treasury has about 110'000 employees.
1422 Assuming every one has an office PC, that is significant computing
1423 power, all of it with plenty of memory for computing "large
1424 memory" hashes. Bot-net operators also have all the memory they
1425 want. The only protection against a resourceful attacker is a
1426 high-entropy passphrase, see items 5.9 and 5.10.
1429 * 5.12 What about iteration count with plain dm-crypt?
1431 Simple: There is none. There is also no salting. If you use plain
1432 dm-crypt, the only way to be secure is to use a high entropy
1433 passphrase. If in doubt, use LUKS instead.
1436 * 5.13 Is LUKS with default parameters less secure on a slow CPU?
1438 Unfortunately, yes. However the only aspect affected is the
1439 protection for low-entropy passphrase or master-key. All other
1440 security aspects are independent of CPU speed.
1442 The master key is less critical, as you really have to work at it
1443 to give it low entropy. One possibility is to supply the master key
1444 yourself. If that key is low-entropy, then you get what you
1445 deserve. The other known possibility is to use /dev/urandom for
1446 key generation in an entropy-starved situation (e.g. automatic
1447 installation on an embedded device without network and other entropy
1450 For the passphrase, don't use a low-entropy passphrase. If your
1451 passphrase is good, then a slow CPU will not matter. If you insist
1452 on a low-entropy passphrase on a slow CPU, use something like
1453 "--iter-time=10" or higher and wait a long time on each LUKS unlock
1454 and pray that the attacker does not find out in which way exactly
1455 your passphrase is low entropy. This also applies to low-entropy
1456 passphrases on fast CPUs. Technology can do only so much to
1457 compensate for problems in front of the keyboard.
1460 * 5.14 Why was the default aes-cbc-plain replaced with aes-cbc-essiv?
1462 Note: This item applies both to plain dm-crypt and to LUKS
1464 The problem is that cbc-plain has a fingerprint vulnerability, where
1465 a specially crafted file placed into the crypto-container can be
1466 recognized from the outside. The issue here is that for cbc-plain
1467 the initialization vector (IV) is the sector number. The IV gets
1468 XORed to the first data chunk of the sector to be encrypted. If you
1469 make sure that the first data block to be stored in a sector
1470 contains the sector number as well, the first data block to be
1471 encrypted is all zeros and always encrypted to the same ciphertext.
1472 This also works if the first data chunk just has a constant XOR
1473 with the sector number. By having several shifted patterns you can
1474 take care of the case of a non-power-of-two start sector number of
1477 This mechanism allows you to create a pattern of sectors that have
1478 the same first ciphertext block and signal one bit per sector to the
1479 outside, allowing you to e.g. mark media files that way for
1480 recognition without decryption. For large files this is a
1481 practical attack. For small ones, you do not have enough blocks to
1482 signal and take care of different file starting offsets.
1484 In order to prevent this attack, the default was changed to
1485 cbc-essiv. ESSIV uses a keyed hash of the sector number, with the
1486 encryption key as key. This makes the IV unpredictable without
1487 knowing the encryption key and the watermarking attack fails.
1490 * 5.15 Are there any problems with "plain" IV? What is "plain64"?
1492 First, "plain" and "plain64" are both not secure to use with CBC,
1493 see previous FAQ item.
1495 However there are modes, like XTS, that are secure with "plain" IV.
1496 The next limit is that "plain" is 64 bit, with the upper 32 bit set
1497 to zero. This means that on volumes larger than 2TiB, the IV
1498 repeats, creating a vulnerability that potentially leaks some
1499 data. To avoid this, use "plain64", which uses the full sector
1500 number up to 64 bit. Note that "plain64" requires a kernel >=
1501 2.6.33. Also note that "plain64" is backwards compatible for
1502 volume sizes <= 2TiB, but not for those > 2TiB. Finally, "plain64"
1503 does not cause any performance penalty compared to "plain".
1506 * 5.16 What about XTS mode?
1508 XTS mode is potentially even more secure than cbc-essiv (but only if
1509 cbc-essiv is insecure in your scenario). It is a NIST standard and
1510 used, e.g. in Truecrypt. From version 1.6.0 of cryptsetup onwards,
1511 aes-xts-plain64 is the default for LUKS. If you want to use it
1512 with a cryptsetup before version 1.6.0 or with plain dm-crypt, you
1513 have to specify it manually as "aes-xts-plain", i.e.
1515 cryptsetup -c aes-xts-plain luksFormat <device>
1517 For volumes >2TiB and kernels >= 2.6.33 use "plain64" (see FAQ
1518 item on "plain" and "plain64"):
1520 cryptsetup -c aes-xts-plain64 luksFormat <device>
1522 There is a potential security issue with XTS mode and large blocks.
1523 LUKS and dm-crypt always use 512B blocks and the issue does not
1527 * 5.17 Is LUKS FIPS-140-2 certified?
1529 No. But that is more a problem of FIPS-140-2 than of LUKS. From a
1530 technical point-of-view, LUKS with the right parameters would be
1531 FIPS-140-2 compliant, but in order to make it certified, somebody
1532 has to pay real money for that. And then, whenever cryptsetup is
1533 changed or extended, the certification lapses and has to be
1536 From the aspect of actual security, LUKS with default parameters
1537 should be as good as most things that are FIPS-140-2 certified,
1538 although you may want to make sure to use /dev/random (by
1539 specifying --use-random on luksFormat) as randomness source for
1540 the master key to avoid being potentially insecure in an
1541 entropy-starved situation.
1544 * 5.18 What about Plausible Deniability?
1546 First let me attempt a definition for the case of encrypted
1547 filesystems: Plausible deniability is when you hide encrypted data
1548 inside an encrypted container and it is not possible to prove it is
1549 there. The idea is compelling and on first glance it seems
1550 possible to do it. And from a cryptographic point of view, it
1551 actually is possible.
1553 So, does it work in practice? No, unfortunately. The reasoning used
1554 by its proponents is fundamentally flawed in several ways and the
1555 cryptographic properties fail fatally when colliding with the real
1558 First, why should "I do not have a hidden partition" be any more
1559 plausible than "I forgot my crypto key" or "I wiped that partition
1560 with random data, nothing in there"? I do not see any reason.
1562 Second, there are two types of situations: Either they cannot force
1563 you to give them the key (then you simply do not) or the can. In
1564 the second case, they can always do bad things to you, because they
1565 cannot prove that you have the key in the first place! This means
1566 they do not have to prove you have the key, or that this random
1567 looking data on your disk is actually encrypted data. So the
1568 situation will allow them to waterboard/lock-up/deport you
1569 anyways, regardless of how "plausible" your deniability is. Do not
1570 have a hidden partition you could show to them, but there are
1571 indications you may? Too bad for you. Unfortunately "plausible
1572 deniability" also means you cannot prove there is no hidden data.
1574 Third, hidden partitions are not that hidden. There are basically
1575 just two possibilities: a) Make a large crypto container, but put a
1576 smaller filesystem in there and put the hidden partition into the
1577 free space. Unfortunately this is glaringly obvious and can be
1578 detected in an automated fashion. This means that the initial
1579 suspicion to put you under duress in order to make you reveal you
1580 hidden data is given. b) Make a filesystem that spans the whole
1581 encrypted partition, and put the hidden partition into space not
1582 currently used by that filesystem. Unfortunately that is also
1583 glaringly obvious, as you then cannot write to the filesystem
1584 without a high risk of destroying data in the hidden container.
1585 Have not written anything to the encrypted filesystem in a while?
1586 Too bad, they have the suspicion they need to do unpleasant things
1589 To be fair, if you prepare option b) carefully and directly before
1590 going into danger, it may work. But then, the mere presence of
1591 encrypted data may already be enough to get you into trouble in
1592 those places were they can demand encryption keys.
1594 Here is an additional reference for some problems with plausible
1595 deniability: http://www.schneier.com/paper-truecrypt-dfs.pdf I
1596 strongly suggest you read it.
1598 So, no, I will not provide any instructions on how to do it with
1599 plain dm-crypt or LUKS. If you insist on shooting yourself in the
1600 foot, you can figure out how to do it yourself.
1603 * 5.19 What about SSDs, Flash and Hybrid Drives?
1605 The problem is that you cannot reliably erase parts of these
1606 devices, mainly due to wear-leveling and possibly defect
1609 Basically, when overwriting a sector (of 512B), what the device
1610 does is to move an internal sector (may be 128kB or even larger) to
1611 some pool of discarded, not-yet erased unused sectors, take a
1612 fresh empty sector from the empty-sector pool and copy the old
1613 sector over with the changes to the small part you wrote. This is
1614 done in some fashion so that larger writes do not cause a lot of
1615 small internal updates.
1617 The thing is that the mappings between outside-addressable sectors
1618 and inside sectors is arbitrary (and the vendors are not talking).
1619 Also the discarded sectors are not necessarily erased immediately.
1620 They may linger a long time.
1622 For plain dm-crypt, the consequences are that older encrypted data
1623 may be lying around in some internal pools of the device. Thus may
1624 or may not be a problem and depends on the application. Remember
1625 the same can happen with a filesystem if consecutive writes to the
1626 same area of a file can go to different sectors.
1628 However, for LUKS, the worst case is that key-slots and LUKS
1629 header may end up in these internal pools. This means that password
1630 management functionality is compromised (the old passwords may
1631 still be around, potentially for a very long time) and that fast
1632 erase by overwriting the header and key-slot area is insecure.
1634 Also keep in mind that the discarded/used pool may be large. For
1635 example, a 240GB SSD has about 16GB of spare area in the chips that
1636 it is free to do with as it likes. You would need to make each
1637 individual key-slot larger than that to allow reliable overwriting.
1638 And that assumes the disk thinks all other space is in use.
1639 Reading the internal pools using forensic tools is not that hard,
1640 but may involve some soldering.
1644 If you trust the device vendor (you probably should not...) you can
1645 try an ATA "secure erase" command for SSDs. That does not work for
1646 USB keys though and may or may not be secure for a hybrid drive. If
1647 it finishes on an SSD after a few seconds, it was possibly faked.
1648 Unfortunately, for hybrid drives that indicator does not work, as
1649 the drive may well take the time to truly erase the magnetic part,
1650 but only mark the SSD/Flash part as erased while data is still in
1653 If you can do without password management and are fine with doing
1654 physical destruction for permanently deleting data (always after
1655 one or several full overwrites!), you can use plain dm-crypt or
1658 If you want or need all the original LUKS security features to work,
1659 you can use a detached LUKS header and put that on a conventional,
1660 magnetic disk. That leaves potentially old encrypted data in the
1661 pools on the disk, but otherwise you get LUKS with the same
1662 security as on a magnetic disk.
1664 If you are concerned about your laptop being stolen, you are likely
1665 fine using LUKS on an SSD or hybrid drive. An attacker would need
1666 to have access to an old passphrase (and the key-slot for this old
1667 passphrase would actually need to still be somewhere in the SSD)
1668 for your data to be at risk. So unless you pasted your old
1669 passphrase all over the Internet or the attacker has knowledge of
1670 it from some other source and does a targeted laptop theft to get
1671 at your data, you should be fine.
1674 * 5.20 LUKS is broken! It uses SHA-1!
1676 No, it is not. SHA-1 is (academically) broken for finding
1677 collisions, but not for using it in a key-derivation function. And
1678 that collision vulnerability is for non-iterated use only. And you
1679 need the hash-value in verbatim.
1681 This basically means that if you already have a slot-key, and you
1682 have set the PBKDF2 iteration count to 1 (it is > 10'000 normally),
1683 you could (maybe) derive a different passphrase that gives you the
1684 the same slot-key. But if you have the slot-key, you can already
1685 unlock the key-slot and get the master key, breaking everything. So
1686 basically, this SHA-1 vulnerability allows you to open a LUKS
1687 container with high effort when you already have it open.
1689 The real problem here is people that do not understand crypto and
1690 claim things are broken just because some mechanism is used that
1691 has been broken for a specific different use. The way the mechanism
1692 is used matters very much. A hash that is broken for one use can be
1693 completely secure for other uses and here it is.
1696 * 5.21 Why is there no "Nuke-Option"?
1698 A "Nuke-Option" or "Kill-switch" is a password that when entered
1699 upon unlocking instead wipes the header and all passwords. So when
1700 somebody forces you to enter your password, you can destroy the
1703 While this sounds attractive at first glance, it does not make sense
1704 once a real security analysis is done. One problem is that you have
1705 to have some kind of HSM (Hardware Security Module) in order to
1706 implement it securely. In the movies, a HSM starts to smoke and
1707 melt once the Nuke-Option has been activated. In reality, it just
1708 wipes some battery-backed RAM cells. A proper HSM costs something
1709 like 20'000...100'000 EUR/USD and there a Nuke-Option may make some
1710 sense. BTW, a chipcard or a TPM is not a HSM, although some
1711 vendors are promoting that myth.
1713 Now, a proper HSMs will have a wipe option but not a Nuke-Option,
1714 i.e. you can explicitly wipe the HSM, but by a different process
1715 than unlocking it takes. Why is that? Simple: If somebody can force
1716 you to reveal passwords, then they can also do bad things to you if
1717 you do not or if you enter a nuke password instead. Think locking
1718 you up for a few years for "destroying evidence" or for far longer
1719 and without trial for being a "terrorist suspect". No HSM maker
1720 will want to expose its customers to that risk.
1722 Now think of the typical LUKS application scenario, i.e. disk
1723 encryption. Usually the ones forcing you to hand over your password
1724 will have access to the disk as well, and, if they have any real
1725 suspicion, they will mirror your disk before entering anything
1726 supplied by you. This neatly negates any Nuke-Option. If they have
1727 no suspicion (just harassing people that cross some border for
1728 example), the Nuke-Option would work, but see above about likely
1729 negative consequences and remember that a Nuke-Option may not work
1730 reliably on SSD and hybrid drives anyways.
1732 Hence my advice is to never take data that you do not want to reveal
1733 into any such situation in the first place. There is no need to
1734 transfer data on physical carriers today. The Internet makes it
1735 quite possible to transfer data between arbitrary places and modern
1736 encryption makes it secure. If you do it right, nobody will even be
1737 able to identify source or destination. (How to do that is out of
1738 scope of this document. It does require advanced skills in this age
1739 of pervasive surveillance.)
1741 Hence, LUKS has not kill option because it would do much more harm
1744 Still, if you have a good use-case (i.e. non-abstract real-world
1745 situation) where a Nuke-Option would actually be beneficial, please
1749 6. Backup and Data Recovery
1752 * 6.1 Why do I need Backup?
1754 First, disks die. The rate for well-treated (!) disk is about 5%
1755 per year, which is high enough to worry about. There is some
1756 indication that this may be even worse for some SSDs. This applies
1757 both to LUKS and plain dm-crypt partitions.
1759 Second, for LUKS, if anything damages the LUKS header or the
1760 key-stripe area then decrypting the LUKS device can become
1761 impossible. This is a frequent occurrence. For example an
1762 accidental format as FAT or some software overwriting the first
1763 sector where it suspects a partition boot sector typically makes a
1764 LUKS partition permanently inaccessible. See more below on LUKS
1767 So, data-backup in some form is non-optional. For LUKS, you may
1768 also want to store a header backup in some secure location. This
1769 only needs an update if you change passphrases.
1772 * 6.2 How do I backup a LUKS header?
1774 While you could just copy the appropriate number of bytes from the
1775 start of the LUKS partition, the best way is to use command option
1776 "luksHeaderBackup" of cryptsetup. This protects also against
1777 errors when non-standard parameters have been used in LUKS
1778 partition creation. Example:
1781 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
1783 To restore, use the inverse command, i.e.
1785 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
1787 If you are unsure about a header to be restored, make a backup of
1788 the current one first! You can also test the header-file without
1789 restoring it by using the --header option for a detached header
1792 cryptsetup --header <file> luksOpen <device> </dev/mapper/ -name>
1794 If that unlocks your keys-lot, you are good. Do not forget to close
1798 * 6.3 How do I test a LUKS header?
1802 cryptsetup -v isLuks <device>
1804 on the device. Without the "-v" it just signals its result via
1805 exit-status. You can also use the more general test
1809 which will also detect other types and give some more info. Omit
1810 "-p" for old versions of blkid that do not support it.
1813 * 6.4 How do I backup a LUKS or dm-crypt partition?
1815 There are two options, a sector-image and a plain file or
1816 filesystem backup of the contents of the partition. The sector
1817 image is already encrypted, but cannot be compressed and contains
1818 all empty space. The filesystem backup can be compressed, can
1819 contain only part of the encrypted device, but needs to be
1820 encrypted separately if so desired.
1822 A sector-image will contain the whole partition in encrypted form,
1823 for LUKS the LUKS header, the keys-slots and the data area. It can
1824 be done under Linux e.g. with dd_rescue (for a direct image copy)
1825 and with "cat" or "dd". Example:
1827 cat /dev/sda10 > sda10.img
1828 dd_rescue /dev/sda10 sda10.img
1830 You can also use any other backup software that is capable of making
1831 a sector image of a partition. Note that compression is
1832 ineffective for encrypted data, hence it does not make sense to
1835 For a filesystem backup, you decrypt and mount the encrypted
1836 partition and back it up as you would a normal filesystem. In this
1837 case the backup is not encrypted, unless your encryption method
1838 does that. For example you can encrypt a backup with "tar" as
1841 tar cjf - <path> | gpg --cipher-algo AES -c - > backup.tbz2.gpg
1843 And verify the backup like this if you are at "path":
1845 cat backup.tbz2.gpg | gpg - | tar djf -
1847 Note: Always verify backups, especially encrypted ones!
1849 There is one problem with verifying like this: The kernel may still
1850 have some files cached and in fact verify them against RAM or may
1851 even verify RAM against RAM, which defeats the purpose of the
1852 exercise. The following command empties the kernel caches:
1854 echo 3 > /proc/sys/vm/drop_caches
1856 Run it after backup and before verify.
1858 In both cases GnuPG will ask you interactively for your symmetric
1859 key. The verify will only output errors. Use "tar dvjf -" to get
1860 all comparison results. To make sure no data is written to disk
1861 unencrypted, turn off swap if it is not encrypted before doing the
1864 Restore works like certification with the 'd' ('difference')
1865 replaced by 'x' ('eXtract'). Refer to the man-page of tar for more
1866 explanations and instructions. Note that with default options tar
1867 will overwrite already existing files without warning. If you are
1868 unsure about how to use tar, experiment with it in a location
1869 where you cannot do damage.
1871 You can of course use different or no compression and you can use
1872 an asymmetric key if you have one and have a backup of the secret
1873 key that belongs to it.
1875 A second option for a filesystem-level backup that can be used when
1876 the backup is also on local disk (e.g. an external USB drive) is
1877 to use a LUKS container there and copy the files to be backed up
1878 between both mounted containers. Also see next item.
1881 * 6.5 Do I need a backup of the full partition? Would the header and
1882 key-slots not be enough?
1884 Backup protects you against two things: Disk loss or corruption
1885 and user error. By far the most questions on the dm-crypt mailing
1886 list about how to recover a damaged LUKS partition are related
1887 to user error. For example, if you create a new filesystem on a
1888 LUKS partition, chances are good that all data is lost
1891 For this case, a header+key-slot backup would often be enough. But
1892 keep in mind that a well-treated (!) HDD has roughly a failure
1893 risk of 5% per year. It is highly advisable to have a complete
1894 backup to protect against this case.
1897 * *6.6 What do I need to backup if I use "decrypt_derived"?
1899 This is a script in Debian, intended for mounting /tmp or swap with
1900 a key derived from the master key of an already decrypted device.
1901 If you use this for an device with data that should be persistent,
1902 you need to make sure you either do not lose access to that master
1903 key or have a backup of the data. If you derive from a LUKS
1904 device, a header backup of that device would cover backing up the
1905 master key. Keep in mind that this does not protect against disk
1908 Note: If you recreate the LUKS header of the device you derive from
1909 (using luksFormat), the master key changes even if you use the same
1910 passphrase(s) and you will not be able to decrypt the derived
1911 device with the new LUKS header.
1914 * 6.7 Does a backup compromise security?
1916 Depends on how you do it. However if you do not have one, you are
1917 going to eventually lose your encrypted data.
1919 There are risks introduced by backups. For example if you
1920 change/disable a key-slot in LUKS, a binary backup of the partition
1921 will still have the old key-slot. To deal with this, you have to
1922 be able to change the key-slot on the backup as well, securely
1923 erase the backup or do a filesystem-level backup instead of a binary
1926 If you use dm-crypt, backup is simpler: As there is no key
1927 management, the main risk is that you cannot wipe the backup when
1928 wiping the original. However wiping the original for dm-crypt
1929 should consist of forgetting the passphrase and that you can do
1930 without actual access to the backup.
1932 In both cases, there is an additional (usually small) risk with
1933 binary backups: An attacker can see how many sectors and which
1934 ones have been changed since the backup. To prevent this, use a
1935 filesystem level backup method that encrypts the whole backup in
1936 one go, e.g. as described above with tar and GnuPG.
1938 My personal advice is to use one USB disk (low value data) or
1939 three disks (high value data) in rotating order for backups, and
1940 either use independent LUKS partitions on them, or use encrypted
1941 backup with tar and GnuPG.
1943 If you do network-backup or tape-backup, I strongly recommend to
1944 go the filesystem backup path with independent encryption, as you
1945 typically cannot reliably delete data in these scenarios,
1946 especially in a cloud setting. (Well, you can burn the tape if it
1947 is under your control...)
1950 * 6.8 What happens if I overwrite the start of a LUKS partition or
1951 damage the LUKS header or key-slots?
1953 There are two critical components for decryption: The salt values
1954 in the key-slot descriptors of the header and the key-slots. If the
1955 salt values are overwritten or changed, nothing (in the
1956 cryptographically strong sense) can be done to access the data,
1957 unless there is a backup of the LUKS header. If a key-slot is
1958 damaged, the data can still be read with a different key-slot, if
1959 there is a remaining undamaged and used key-slot. Note that in
1960 order to make a key-slot unrecoverable in a cryptographically
1961 strong sense, changing about 4-6 bits in random locations of its
1962 128kiB size is quite enough.
1965 * 6.9 What happens if I (quick) format a LUKS partition?
1967 I have not tried the different ways to do this, but very likely you
1968 will have written a new boot-sector, which in turn overwrites the
1969 LUKS header, including the salts, making your data permanently
1970 irretrievable, unless you have a LUKS header backup. You may also
1971 damage the key-slots in part or in full. See also last item.
1974 * 6.10 How do I recover the master key from a mapped LUKS container?
1976 This is typically only needed if you managed to damage your LUKS
1977 header, but the container is still mapped, i.e. "luksOpen"ed. It
1978 also helps if you have a mapped container that you forgot or do not
1979 know a passphrase for (e.g. on a long running server.)
1981 WARNING: Things go wrong, do a full backup before trying this!
1983 WARNING: This exposes the master key of the LUKS container. Note
1984 that both ways to recreate a LUKS header with the old master key
1985 described below will write the master key to disk. Unless you are
1986 sure you have securely erased it afterwards, e.g. by writing it to
1987 an encrypted partition, RAM disk or by erasing the filesystem you
1988 wrote it to by a complete overwrite, you should change the master
1989 key afterwards. Changing the master key requires a full data
1990 backup, luksFormat and then restore of the backup.
1992 First, there is a script by Milan that automates the whole
1993 process, except generating a new LUKS header with the old master
1994 key (it prints the command for that though):
1996 https://gitlab.com/cryptsetup/cryptsetup/blob/master/misc/luks-header-from-active
1998 You can also do this manually. Here is how:
2000 - Get the master key from the device mapper. This is done by the
2001 following command. Substitute c5 for whatever you mapped to:
2003 # dmsetup table --target crypt --showkey /dev/mapper/c5
2005 0 200704 crypt aes-cbc-essiv:sha256
2006 a1704d9715f73a1bb4db581dcacadaf405e700d591e93e2eaade13ba653d0d09
2009 The result is actually one line, wrapped here for clarity. The long
2010 hex string is the master key.
2012 - Convert the master key to a binary file representation. You can
2013 do this manually, e.g. with hexedit. You can also use the tool
2014 "xxd" from vim like this:
2016 echo "a1704d9....53d0d09" | xxd -r -p > <master-key-file>
2018 - Do a luksFormat to create a new LUKS header.
2020 NOTE: If your header is intact and you just forgot the
2021 passphrase, you can just set a new passphrase, see next
2024 Unmap the device before you do that (luksClose). Then do
2026 cryptsetup luksFormat --master-key-file=<master-key-file> <luks device>
2028 Note that if the container was created with other than the default
2029 settings of the cryptsetup version you are using, you need to give
2030 additional parameters specifying the deviations. If in doubt, try
2031 the script by Milan. It does recover the other parameters as well.
2033 Side note: This is the way the decrypt_derived script gets at the
2034 master key. It just omits the conversion and hashes the master key
2037 - If the header is intact and you just forgot the passphrase, just
2038 set a new passphrase like this:
2040 cryptsetup luksAddKey --master-key-file=<master-key-file> <luks device>
2042 You may want to disable the old one afterwards.
2045 * 6.11 What does the on-disk structure of dm-crypt look like?
2047 There is none. dm-crypt takes a block device and gives encrypted
2048 access to each of its blocks with a key derived from the passphrase
2049 given. If you use a cipher different than the default, you have to
2050 specify that as a parameter to cryptsetup too. If you want to
2051 change the password, you basically have to create a second
2052 encrypted device with the new passphrase and copy your data over.
2053 On the plus side, if you accidentally overwrite any part of a
2054 dm-crypt device, the damage will be limited to the area you
2058 * 6.12 What does the on-disk structure of LUKS look like?
2060 A LUKS partition consists of a header, followed by 8 key-slot
2061 descriptors, followed by 8 key slots, followed by the encrypted
2064 Header and key-slot descriptors fill the first 592 bytes. The
2065 key-slot size depends on the creation parameters, namely on the
2066 number of anti-forensic stripes, key material offset and master
2069 With the default parameters, each key-slot is a bit less than
2070 128kiB in size. Due to sector alignment of the key-slot start,
2071 that means the key block 0 is at offset 0x1000-0x20400, key
2072 block 1 at offset 0x21000-0x40400, and key block 7 at offset
2073 0xc1000-0xe0400. The space to the next full sector address is
2074 padded with zeros. Never used key-slots are filled with what the
2075 disk originally contained there, a key-slot removed with
2076 "luksRemoveKey" or "luksKillSlot" gets filled with 0xff. Due to
2077 2MiB default alignment, start of the data area for cryptsetup 1.3
2078 and later is at 2MiB, i.e. at 0x200000. For older versions, it is
2079 at 0x101000, i.e. at 1'052'672 bytes, i.e. at 1MiB + 4096 bytes
2080 from the start of the partition. Incidentally, "luksHeaderBackup"
2081 for a LUKS container created with default parameters dumps exactly
2082 the first 2MiB (or 1'052'672 bytes for headers created with
2083 cryptsetup versions < 1.3) to file and "luksHeaderRestore" restores
2086 For non-default parameters, you have to figure out placement
2087 yourself. "luksDump" helps. See also next item. For the most common
2088 non-default settings, namely aes-xts-plain with 512 bit key, the
2089 offsets are: 1st keyslot 0x1000-0x3f800, 2nd keyslot
2090 0x40000-0x7e000, 3rd keyslot 0x7e000-0xbd800, ..., and start of
2091 bulk data at 0x200000.
2093 The exact specification of the format is here:
2094 https://gitlab.com/cryptsetup/cryptsetup/wikis/Specification
2096 For your convenience, here is the LUKS header with hex offsets.
2097 NOTE: The spec counts key-slots from 1 to 8, but the cryptsetup
2098 tool counts from 0 to 7. The numbers here refer to the cryptsetup
2101 Refers to LUKS On-Disk Format Specification Version 1.2.1
2103 offset length name data type description
2104 -----------------------------------------------------------------------
2105 0x0000 0x06 magic byte[] 'L','U','K','S', 0xba, 0xbe
2107 0x0006 0x02 version uint16_t LUKS version
2109 0x0008 0x20 cipher-name char[] cipher name spec.
2111 0x0028 0x20 cipher-mode char[] cipher mode spec.
2113 0x0048 0x20 hash-spec char[] hash spec.
2115 0x0068 0x04 payload-offset uint32_t bulk data offset in sectors
2116 104 4 (512 bytes per sector)
2117 0x006c 0x04 key-bytes uint32_t number of bytes in key
2119 0x0070 0x14 mk-digest byte[] master key checksum
2120 112 20 calculated with PBKDF2
2121 0x0084 0x20 mk-digest-salt byte[] salt for PBKDF2 when
2122 132 32 calculating mk-digest
2123 0x00a4 0x04 mk-digest-iter uint32_t iteration count for PBKDF2
2124 164 4 when calculating mk-digest
2125 0x00a8 0x28 uuid char[] partition UUID
2127 0x00d0 0x30 key-slot-0 key slot key slot 0
2129 0x0100 0x30 key-slot-1 key slot key slot 1
2131 0x0130 0x30 key-slot-2 key slot key slot 2
2133 0x0160 0x30 key-slot-3 key slot key slot 3
2135 0x0190 0x30 key-slot-4 key slot key slot 4
2137 0x01c0 0x30 key-slot-5 key slot key slot 5
2139 0x01f0 0x30 key-slot-6 key slot key slot 6
2141 0x0220 0x30 key-slot-7 key slot key slot 7
2144 offset length name data type description
2145 -------------------------------------------------------------------------
2146 0x0000 0x04 active uint32_t key slot enabled/disabled
2148 0x0004 0x04 iterations uint32_t PBKDF2 iteration count
2150 0x0008 0x20 salt byte[] PBKDF2 salt
2152 0x0028 0x04 key-material-offset uint32_t key start sector
2153 40 4 (512 bytes/sector)
2154 0x002c 0x04 stripes uint32_t number of anti-forensic
2158 * 6.13 What is the smallest possible LUKS container?
2160 Note: From cryptsetup 1.3 onwards, alignment is set to 1MB. With
2161 modern Linux partitioning tools that also align to 1MB, this will
2162 result in alignment to 2k sectors and typical Flash/SSD sectors,
2163 which is highly desirable for a number of reasons. Changing the
2164 alignment is not recommended.
2166 That said, with default parameters, the data area starts at
2167 exactly 2MB offset (at 0x101000 for cryptsetup versions before
2168 1.3). The smallest data area you can have is one sector of 512
2169 bytes. Data areas of 0 bytes can be created, but fail on mapping.
2171 While you cannot put a filesystem into something this small, it may
2172 still be used to contain, for example, key. Note that with current
2173 formatting tools, a partition for a container this size will be
2174 3MiB anyways. If you put the LUKS container into a file (via
2175 losetup and a loopback device), the file needs to be 2097664 bytes
2176 in size, i.e. 2MiB + 512B.
2178 There two ways to influence the start of the data area are key-size
2181 For alignment, you can go down to 1 on the parameter. This will
2182 still leave you with a data-area starting at 0x101000, i.e.
2183 1MiB+4096B (default parameters) as alignment will be rounded up to
2184 the next multiple of 8 (i.e. 4096 bytes) If in doubt, do a dry-run
2185 on a larger file and dump the LUKS header to get actual
2188 For key-size, you can use 128 bit (e.g. AES-128 with CBC), 256 bit
2189 (e.g. AES-256 with CBC) or 512 bit (e.g. AES-256 with XTS mode).
2190 You can do 64 bit (e.g. blowfish-64 with CBC), but anything below
2191 128 bit has to be considered insecure today.
2193 Example 1 - AES 128 bit with CBC:
2195 cryptsetup luksFormat -s 128 --align-payload=8 <device>
2197 This results in a data offset of 0x81000, i.e. 516KiB or 528384
2198 bytes. Add one 512 byte sector and the smallest LUKS container size
2199 with these parameters is 516KiB + 512B or 528896 bytes.
2201 Example 2 - Blowfish 64 bit with CBC (WARNING: insecure):
2203 cryptsetup luksFormat -c blowfish -s 64 --align-payload=8 /dev/loop0
2205 This results in a data offset of 0x41000, i.e. 260kiB or 266240
2206 bytes, with a minimal LUKS container size of 260kiB + 512B or
2210 * 6.14 I think this is overly complicated. Is there an alternative?
2212 Not really. Encryption comes at a price. You can use plain
2213 dm-crypt to simplify things a bit. It does not allow multiple
2214 passphrases, but on the plus side, it has zero on disk description
2215 and if you overwrite some part of a plain dm-crypt partition,
2216 exactly the overwritten parts are lost (rounded up to sector
2220 * 6.15 Can I clone a LUKS container?
2222 You can, but it breaks security, because the cloned container has
2223 the same header and hence the same master key. You cannot change
2224 the master key on a LUKS container, even if you change the
2225 passphrase(s), the master key stays the same. That means whoever
2226 has access to one of the clones can decrypt them all, completely
2227 bypassing the passphrases.
2229 The right way to do this is to first luksFormat the target
2230 container, then to clone the contents of the source container, with
2231 both containers mapped, i.e. decrypted. You can clone the decrypted
2232 contents of a LUKS container in binary mode, although you may run
2233 into secondary issues with GUIDs in filesystems, partition tables,
2234 RAID-components and the like. These are just the normal problems
2235 binary cloning causes.
2237 Note that if you need to ship (e.g.) cloned LUKS containers with a
2238 default passphrase, that is fine as long as each container was
2239 individually created (and hence has its own master key). In this
2240 case, changing the default passphrase will make it secure again.
2243 7. Interoperability with other Disk Encryption Tools
2246 * 7.1 What is this section about?
2248 Cryptsetup for plain dm-crypt can be used to access a number of
2249 on-disk formats created by tools like loop-aes patched into
2250 losetup. This sometimes works and sometimes does not. This
2251 section collects insights into what works, what does not and where
2252 more information is required.
2254 Additional information may be found in the mailing-list archives,
2255 mentioned at the start of this FAQ document. If you have a
2256 solution working that is not yet documented here and think a wider
2257 audience may be interested, please email the FAQ maintainer.
2260 * 7.2 loop-aes: General observations.
2262 One problem is that there are different versions of losetup around.
2263 loop-aes is a patch for losetup. Possible problems and deviations
2264 from cryptsetup option syntax include:
2266 - Offsets specified in bytes (cryptsetup: 512 byte sectors)
2268 - The need to specify an IV offset
2270 - Encryption mode needs specifying (e.g. "-c twofish-cbc-plain")
2272 - Key size needs specifying (e.g. "-s 128" for 128 bit keys)
2274 - Passphrase hash algorithm needs specifying
2276 Also note that because plain dm-crypt and loop-aes format does not
2277 have metadata, and while the loopAES extension for cryptsetup tries
2278 autodetection (see command loopaesOpen), it may not always work.
2279 If you still have the old set-up, using a verbosity option (-v)
2280 on mapping with the old tool or having a look into the system logs
2281 after setup could give you the information you need. Below, there
2282 are also some things that worked for somebody.
2285 * 7.3 loop-aes patched into losetup on Debian 5.x, kernel 2.6.32
2287 In this case, the main problem seems to be that this variant of
2288 losetup takes the offset (-o option) in bytes, while cryptsetup
2289 takes it in sectors of 512 bytes each. Example: The losetup command
2291 losetup -e twofish -o 2560 /dev/loop0 /dev/sdb1
2292 mount /dev/loop0 mount-point
2296 cryptsetup create -c twofish -o 5 --skip 5 e1 /dev/sdb1
2297 mount /dev/mapper/e1 mount-point
2300 * 7.4 loop-aes with 160 bit key
2302 This seems to be sometimes used with twofish and blowfish and
2303 represents a 160 bit ripemed160 hash output padded to 196 bit key
2304 length. It seems the corresponding options for cryptsetup are
2306 --cipher twofish-cbc-null -s 192 -h ripemd160:20
2309 * 7.5 loop-aes v1 format OpenSUSE
2311 Apparently this is done by older OpenSUSE distros and stopped
2312 working from OpenSUSE 12.1 to 12.2. One user had success with the
2315 cryptsetup create <target> <device> -c aes -s 128 -h sha256
2318 * 7.6 Kernel encrypted loop device (cryptoloop)
2320 There are a number of different losetup implementations for using
2321 encrypted loop devices so getting this to work may need a bit of
2324 NOTE: Do NOT use this for new containers! Some of the existing
2325 implementations are insecure and future support is uncertain.
2327 Example for a compatible mapping:
2329 losetup -e twofish -N /dev/loop0 /image.img
2333 cryptsetup create image_plain /image.img -c twofish-cbc-plain -H plain
2335 with the mapping being done to /dev/mapper/image_plain instead of
2340 Cipher, mode and pasword hash (or no hash):
2342 -e cipher [-N] => -c cipher-cbc-plain -H plain [-s 256]
2343 -e cipher => -c cipher-cbc-plain -H ripemd160 [-s 256]
2345 Key size and offsets (losetup: bytes, cryptsetuop: sectors of 512
2349 -o 2560 => -o 5 -p 5 # 2560/512 = 5
2351 There is no replacement for --pass-fd, it has to be emulated using
2352 keyfiles, see the cryptsetup man-page.
2355 8. Issues with Specific Versions of cryptsetup
2358 * 8.1 When using the create command for plain dm-crypt with
2359 cryptsetup 1.1.x, the mapping is incompatible and my data is not
2362 With cryptsetup 1.1.x, the distro maintainer can define different
2363 default encryption modes. You can check the compiled-in defaults
2364 using "cryptsetup --help". Moreover, the plain device default
2365 changed because the old IV mode was vulnerable to a watermarking
2368 If you are using a plain device and you need a compatible mode, just
2369 specify cipher, key size and hash algorithm explicitly. For
2370 compatibility with cryptsetup 1.0.x defaults, simple use the
2373 cryptsetup create -c aes-cbc-plain -s 256 -h ripemd160 <name> <dev>
2375 LUKS stores cipher and mode in the metadata on disk, avoiding this
2379 * 8.2 cryptsetup on SLED 10 has problems...
2381 SLED 10 is missing an essential kernel patch for dm-crypt, which
2382 is broken in its kernel as a result. There may be a very old
2383 version of cryptsetup (1.0.x) provided by SLED, which should also
2384 not be used anymore as well. My advice would be to drop SLED 10.
2387 * 8.3 Gcrypt after 1.5.3 breaks Whirlpool
2389 It is the other way round: In gcrypt 1.5.3 and before Whirlpool is
2390 broken and it was fixed in the next version. If you selected
2391 whirlpool as hash on creation of a LUKS container, it does not work
2392 anymore with the fixed library. This shows one serious risk of
2393 using rarely used settings.
2395 The only two ways to deal with this are either to decrypt with an
2396 old gcrypt version that has the flaw or to use a compatibility
2397 feature introduced in cryptsetup 1.6.4 and gcrypt 1.6.1 or later.
2398 Versions of gcrypt between 1.5.4 and 1.6.0 cannot be used.
2402 - Make a least a header backup or better, refresh your full
2403 backup. (You have a full backup, right? See Item 6.1 and
2406 - Make sure you have cryptsetup 1.6.4 or later and check the gcrypt
2410 cryptsetup luksDump <your luks device> --debug | grep backend
2412 If gcrypt is at version 1.5.3 or before:
2414 - Reencrypt the LUKS header with a different hash. (Requires
2415 entering all keyslot passphrases. If you do not have all, remove
2416 the ones you do not have before.):
2418 cryptsetup-reencrypt --keep-key --hash sha256 <your luks device>
2420 If gcrypt is at version 1.6.1 or later:
2422 - Patch the hash name in the LUKS header from "whirlpool" to
2423 "whirlpool_gcryptbug". This activates the broken implementation.
2424 The detailed header layout is in Item 6.12 of this FAQ and in the
2425 LUKS on-disk format specification. One way to change the hash is
2426 with the following command:
2428 echo -n -e 'whirlpool_gcryptbug\0' | dd of=<luks device> bs=1 seek=72 conv=notrunc
2430 - You can now open the device again. It is highly advisable to
2431 change the hash now with cryptsetup-reencrypt as described above.
2432 While you can reencrypt to use the fixed whirlpool, that may not
2433 be a good idea as almost nobody seems to use it and hence the long
2434 time until the bug was discovered.
2437 9. References and Further Reading
2440 * Purpose of this Section
2442 The purpose of this section is to collect references to all
2443 materials that do not fit the FAQ but are relevant in some fashion.
2444 This can be core topics like the LUKS spec or disk encryption, but
2445 it can also be more tangential, like secure storage management or
2446 cryptography used in LUKS. It should still have relevance to
2447 cryptsetup and its applications.
2449 If you wan to see something added here, send email to the
2450 maintainer (or the cryptsetup mailing list) giving an URL, a
2451 description (1-3 lines preferred) and a section to put it in. You
2452 can also propose new sections.
2454 At this time I would like to limit the references to things that
2455 are available on the web.
2460 - LUKS on-disk format spec:
2461 https://gitlab.com/cryptsetup/cryptsetup/wikis/Specification
2465 - Some code examples are in the source package under docs/examples
2468 * Brute-forcing passphrases
2471 http://news.electricalchemy.net/2009/10/password-cracking-in-cloud-part-5.html
2474 http://it.slashdot.org/story/12/12/05/0623215/new-25-gpu-monster-devours-strong-passwords-in-minutes
2480 * SSD and Flash Disk Related
2486 * Attacks Against Disk Encryption
2489 * Risk Management as Relevant for Disk Encryption
2497 A. Contributors In no particular order: