8 6. Backup and Data Recovery
9 7. Interoperability with other Disk Encryption Tools
10 8. Issues with Specific Versions of cryptsetup
19 This is the FAQ (Frequently Asked Questions) for cryptsetup. It
20 covers Linux disk encryption with plain dm-crypt (one passphrase,
21 no management, no metadata on disk) and LUKS (multiple user keys
22 with one master key, anti-forensic features, metadata block at
23 start of device, ...). The latest version of this FAQ should
24 usually be available at
25 http://code.google.com/p/cryptsetup/wiki/FrequentlyAskedQuestions
30 ATTENTION: If you are going to read just one thing, make it the
31 section on Backup and Data Recovery. By far the most questions on
32 the cryptsetup mailing list are from people that managed to damage
33 the start of their LUKS partitions, i.e. the LUKS header. In
34 most cases, there is nothing that can be done to help these poor
35 souls recover their data. Make sure you understand the problem and
36 limitations imposed by the LUKS security model BEFORE you face
37 such a disaster! In particular, make sure you have a current header
38 backup before doing any potentially dangerous operations.
40 DISTRIBUTION INSTALLERS: Some distribution installers offer to
41 create LUKS containers in a way that can be mistaken as activation
42 of an existing container. Creating a new LUKS container on top of
43 an existing one leads to permanent, complete and irreversible data
44 loss. It is strongly recommended to only use distribution
45 installers after a complete backup of all LUKS containers has been
48 LUKS PASSPHRASE IS NOT THE MASTER KEY: The LUKS passphrase is not
49 used in deriving the master key. It is used in decrypting a master
50 key that is randomly selected on header creation. This means that
51 if you create a new LUKS header on top of an old one with
52 exactly the same parameters and exactly the same passphrase as the
53 old one, it will still have a different master key and your data
54 will be permanently lost.
56 PASSPHRASE CHARACTER SET: Some people have had difficulties with
57 this when upgrading distributions. It is highly advisable to only
58 use the 94 printable characters from the first 128 characters of
59 the ASCII table, as they will always have the same binary
60 representation. Other characters may have different encoding
61 depending on system configuration and your passphrase will not
62 work with a different encoding. A table of the standardized first
63 128 ASCII caracters can, e.g. be found on
64 http://en.wikipedia.org/wiki/ASCII
67 * System Specific warnings
69 - Ubuntu as of 4/2011: It seems the installer offers to create
70 LUKS partitions in a way that several people mistook for an offer
71 to activate their existing LUKS partition. The installer gives no
72 or an inadequate warning and will destroy your old LUKS header,
73 causing permanent data loss. See also the section on Backup and
76 This issue has been acknowledged by the Ubuntu dev team, see here:
77 http://launchpad.net/bugs/420080
82 Current FAQ maintainer is Arno Wagner <arno@wagner.name>. Other
83 contributors are listed at the end. If you want to contribute, send
84 your article, including a descriptive headline, to the maintainer,
85 or the dm-crypt mailing list with something like "FAQ ..." in the
86 subject. You can also send more raw information and have me write
87 the section. Please note that by contributing to this FAQ, you
88 accept the license described below.
90 This work is under the "Attribution-Share Alike 3.0 Unported"
91 license, which means distribution is unlimited, you may create
92 derived works, but attributions to original authors and this
93 license statement must be retained and the derived work must be
94 under the same license. See
95 http://creativecommons.org/licenses/by-sa/3.0/ for more details of
98 Side note: I did text license research some time ago and I think
99 this license is best suited for the purpose at hand and creates the
103 * Where is the project website?
105 There is the project website at http://code.google.com/p/cryptsetup/
106 Please do not post questions there, nobody will read them. Use
107 the mailing-list instead.
110 * Is there a mailing-list?
112 Instructions on how to subscribe to the mailing-list are at on the
113 project website. People are generally helpful and friendly on the
116 The question of how to unsubscribe from the list does crop up
117 sometimes. For this you need your list management URL, which is
118 sent to you initially and once at the start of each month. Go to
119 the URL mentioned in the email and select "unsubscribe". This page
120 also allows you to request a password reminder.
122 Alternatively, you can send an Email to dm-crypt-request@saout.de
123 with just the word "help" in the subject or message body. Make sure
124 to send it from your list address.
126 The mailing list archive is here:
127 http://dir.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt
133 * What is the difference between "plain" and LUKS format?
135 Plain format is just that: It has no metadata on disk, reads all
136 paramters from the commandline (or the defaults), derives a
137 master-key from the passphrase and then uses that to de-/encrypt
138 the sectors of the device, with a direct 1:1 mapping between
139 encrypted and decrypted sectors.
141 Primary advantage is high resilience to damage, as one damaged
142 encrypted sector results in exactly one damaged decrypted sector.
143 Also, it is not readily apparent that there even is encrypted data
144 on the device, as an overwrite with crypto-grade randomness (e.g.
145 from /dev/urandom) looks exactly the same on disk.
147 Side-note: That has limited value against the authorities. In
148 civilized countries, they cannot force you to give up a crypto-key
149 anyways. In the US, the UK and dictatorships around the world,
150 they can force you to give up the keys (using imprisonment or worse
151 to pressure you), and in the worst case, they only need a
152 nebulous "suspicion" about the presence of encrypted data. My
153 advice is to either be ready to give up the keys or to not have
154 encrypted data when traveling to those countries, especially when
155 crossing the borders.
157 Disadvantages are that you do not have all the nice features that
158 the LUKS metadata offers, like multiple passphrases that can be
159 changed, the cipher being stored in the metadata, anti-forensic
160 properties like key-slot diffusion and salts, etc..
162 LUKS format uses a metadata header and 8 key-slot areas that are
163 being placed ath the begining of the disk, see below under "What
164 does the LUKS on-disk format looks like?". The passphrases are used
165 to decryt a single master key that is stored in the anti-forensic
168 Advantages are a higher usability, automatic configuration of
169 non-default crypto parameters, defenses against low-entropy
170 passphrases like salting and iterated PBKDF2 passphrase hashing,
171 the ability to change passhrases, and others.
173 Disadvantages are that it is readily obvious there is encrypted
174 data on disk (but see side note above) and that damage to the
175 header or key-slots usually results in permanent data-loss. See
176 below under "6. Backup and Data Recovery" on how to reduce that
177 risk. Also the sector numbers get shifted by the length of the
178 header and key-slots and there is a loss of that size in capacity
179 (1MB+4096B for defaults and 2MB for the most commonly used
180 non-default XTS mode).
183 * Can I encrypt an already existing, non-empty partition to use LUKS?
185 There is no converter, and it is not really needed. The way to do
186 this is to make a backup of the device in question, securely wipe
187 the device (as LUKS device initialization does not clear away old
188 data), do a luksFormat, optionally overwrite the encrypted device,
189 create a new filesystem and restore your backup on the now
190 encrypted device. Also refer to sections "Security Aspects" and
191 "Backup and Data Recovery".
193 For backup, plain GNU tar works well and backs up anything likely
194 to be in a filesystem.
197 * How do I use LUKS with a loop-device?
199 This can be very handy for experiments. Setup is just the same as
200 with any block device. If you want, for example, to use a 100MiB
201 file as LUKS container, do something like this:
203 head -c 100M /dev/zero > luksfile # create empty file
204 losetup /dev/loop0 luksfile # map luksfile to /dev/loop0
205 cryptsetup luksFormat /dev/loop0 # create LUKS on loop device
207 Afterwards just use /dev/loop0 as a you would use a LUKS partition.
208 To unmap the file when done, use "losetup -d /dev/loop0".
211 * When I add a new key-slot to LUKS, it asks for a passphrase but
212 then complains about there not being a key-slot with that
215 That is as intended. You are asked a passphrase of an existing
216 key-slot first, before you can enter the passphrase for the new
217 key-slot. Otherwise you could break the encryption by just adding a
218 new key-slot. This way, you have to know the passphrase of one of
219 the already configured key-slots in order to be able to configure a
223 * Encrytion on top of RAID or the other way round?
225 Unless you have special needs, place encryption between RAID and
226 filesystem, i.e. encryption on top of RAID. You can do it the other
227 way round, but you have to be aware that you then need to give the
228 pasphrase for each individual disk and RAID autotetection will not
229 work anymore. Therefore it is better to encrypt the RAID device,
233 * How do I read a dm-crypt key from file?
235 Note that the file will still be hashed first, just like keyboard
236 input. Use the --key-file option, like this:
238 cryptsetup create --key-file keyfile e1 /dev/loop0
241 * How do I read a LUKS slot key from file?
243 What you really do here is to read a passphrase from file, just as
244 you would with manual entry of a passphrase for a key-slot. You can
245 add a new passphrase to a free key-slot, set the passphrase of an
246 specific key-slot or put an already configured passphrase into a
247 file. In the last case make sure no trailing newline (0x0a) is
248 contained in the key file, or the passphrase will not work because
249 the whole file is used as input.
251 To add a new passphrase to a free key slot from file, use something
254 cryptsetup luksAddKey /dev/loop0 keyfile
256 To add a new passphrase to a specific key-slot, use something like
259 cryptsetup luksAddKey --key-slot 7 /dev/loop0 keyfile
261 To supply a key from file to any LUKS command, use the --key-file
262 option, e.g. like this:
264 cryptsetup luksOpen --key-file keyfile /dev/loop0 e1
267 * How do I read the LUKS master key from file?
269 The question you should ask yourself first is why you would want to
270 do this. The only legitimate reason I can think of is if you want
271 to have two LUKS devices with the same master key. Even then, I
272 think it would be preferable to just use key-slots with the same
273 passphrase, or to use plain dm-crypt instead. If you really have a
274 good reason, please tell me. If I am convinced, I will add how to
278 * What are the security requirements for a key read from file?
280 A file-stored key or passphrase has the same security requirements
281 as one entered interactively, however you can use random bytes and
282 thereby use bytes you cannot type on the keyboard. You can use any
283 file you like as key file, for example a plain text file with a
284 human readable passphrase. To generate a file with random bytes,
285 use something like this:
287 head -c 256 /dev/random > keyfile
290 * If I map a journaled file system using dm-crypt/LUKS, does it still
291 provide its usual transactional guarantees?
293 As far as I know it does (but I may be wrong), but please note that
294 these "guarantees" are far weaker than they appear to be. For
295 example, you may not get a hard flush to disk surface even on a
296 call to fsync. In addition, the HDD itself may do independent
297 write reordering. Some other things can go wrong as well. The
298 filesystem developers are aware of these problems and typically
299 can make it work anyways. That said, dm-crypt/LUKS should not make
302 Personally, I have several instances of ext3 on dm-crypt and have
303 not noticed any specific problems.
305 Update: I did run into frequent small freezes (1-2 sec) when putting
306 a vmware image on ext3 over dm-crypt. This does indicate that the
307 transactional guarantees are in place, but at a cost. When I went
308 back to ext2, the problem went away. This also seems to have gotten
309 better with kernel 2.6.36 and the reworking of filesystem flush
310 locking. Kernel 2.6.38 is expected to have more improvements here.
313 * Can I use LUKS or cryptsetup with a more secure (external) medium
314 for key storage, e.g. TPM or a smartcard?
316 Yes, see the answers on using a file-supplied key. You do have to
317 write the glue-logic yourself though. Basically you can have
318 cryptsetup read the key from STDIN and write it there with your
319 own tool that in turn gets the key from the more secure key
323 * Can I resize a dm-crypt or LUKS partition?
325 Yes, you can, as neither dm-crypt nor LUKS stores partition size.
326 Whether you should is a different question. Personally I recommend
327 backup, recreation of the encrypted partition with new size,
328 recreation of the filesystem and restore. This gets around the
329 tricky business of resizing the filesystem. Resizing a dm-crypt or
330 LUKS container does not resize the filesystem in it. The backup is
331 really non-optional here, as a lot can go wrong, resulting in
332 partial or complete data loss. Using something like gparted to
333 resize an encrypted partition is slow, but typicaly works. This
334 will not change the size of the filesystem hidden under the
337 You also need to be aware of size-based limitations. The one
338 currently relevant is that aes-xts-plain should not be used for
339 encrypted container sizes larger than 2TiB. Use aes-xts-plain64
346 * My dm-crypt/LUKS mapping does not work! What general steps are
347 there to investigate the problem?
349 If you get a specific error message, investigate what it claims
350 first. If not, you may want to check the following things.
352 - Check that "/dev", including "/dev/mapper/control" is there. If it
353 is missing, you may have a problem with the "/dev" tree itself or
354 you may have broken udev rules.
356 - Check that you have the device mapper and the crypt target in your
357 kernel. The output of "dmsetup targets" should list a "crypt"
358 target. If it is not there or the command fails, add device mapper
359 and crypt-target to the kernel.
361 - Check that the hash-functions and ciphers you want to use are in
362 the kernel. The output of "cat /proc/crypto" needs to list them.
365 * My dm-crypt mapping suddenly stopped when upgrading cryptsetup.
367 The default cipher, hash or mode may have changed (the mode changed
368 from 1.0.x to 1.1.x). See under "Issues With Specific Versions of
372 * When I call cryptsetup from cron/CGI, I get errors about unknown
375 If you get errors about unknown parameters or the like that are not
376 present when cryptsetup is called from the shell, make sure you
377 have no older version of cryptsetup on your system that then gets
378 called by cron/CGI. For example some distributions install
379 cryptsetup into /usr/sbin, while a manual install could go to
380 /usr/local/sbin. As a debugging aid, call "cryptsetup --version"
381 from cron/CGI or the non-shell mechanism to be sure the right
385 * Unlocking a LUKS device takes very long. Why?
387 The iteration time for a key-slot (see Section 5 for an explanation
388 what iteration does) is calculated when setting a passphrase. By
389 default it is 1 second on the machine where the passphrase is set.
390 If you set a passphrase on a fast machine and then unlock it on a
391 slow machine, the unlocking time can be much longer. Also take into
392 account that up to 8 key-slots have to be tried in order to find the
395 If this is problem, you can add another key-slot using the slow
396 machine with the same passphrase and then remove the old key-slot.
397 The new key-slot will have an iteration count adjusted to 1 second
398 on the slow machine. Use luksKeyAdd and then luksKillSlot or
401 However, this operation will not change volume key iteration count
402 (MK iterations in output of "cryptsetup luksDump"). In order to
403 change that, you will have to backup the data in the LUKS
404 container (i.e. your encrypted data), luksFormat on the slow
405 machine and restore the data. Note that in the original LUKS
406 specification this value was fixed to 10, but it is now derived
407 from the PBKDF2 benchmark as well and set to iterations in 0.125
408 sec or 1000, whichever is larger. Also note that MK iterations
409 are not very security relevant. But as each key-slot already takes
410 1 second, spending the additional 0.125 seconds really does not
414 * "blkid" sees a LUKS UUID and an ext2/swap UUID on the same device.
417 Some old versions of cryptsetup have a bug where the header does
418 not get completely wiped during LUKS format and an older ext2/swap
419 signature remains on the device. This confuses blkid.
421 Fix: Wipe the unused header areas by doing a backup and restore of
422 the header with cryptsetup 1.1.x:
424 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
425 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
428 * cryptsetup segfaults on Gentoo amd64 hardened ...
430 There seems to be some inteference between the hardening and and
431 the way cryptsetup benchmarks PBKDF2. The solution to this is
432 currently not quite clear for an encrypted root filesystem. For
433 other uses, you can apparently specify USE="dynamic" as compile
434 flag, see http://bugs.gentoo.org/show_bug.cgi?id=283470
440 * Can a bad RAM module cause problems?
442 LUKS and dm-crypt can give the RAM quite a workout, especially when
443 combined with software RAID. In particular the combination RAID5 +
444 LUKS + XFS seems to uncover RAM problems that never caused obvious
445 problems before. Symptoms vary, but often the problem manifest
446 itself when copying large amounts of data, typically several times
447 larger than your main memory.
449 Side note: One thing you should always do on large data
450 copy/movements is to run a verify, for example with the "-d"
451 option of "tar" or by doing a set of MD5 checksums on the source
454 find . -type f -exec md5sum \{\} \; > checksum-file
456 and then a "md5sum -c checksum-file" on the other side. If you get
457 mismatches here, RAM is the primary suspect. A lesser suspect is
458 an overclocked CPU. I have found countless hardware problems in
459 verify runs after copying or making backups. Bit errors are much
460 more common than most people think.
462 Some RAM issues are even worse and corrupt structures in one of the
463 layers. This typically results in lockups, CPU state dumps in the
464 system logs, kernel panic or other things. It is quite possible to
465 have the problem with an encrypted device, but not with an
466 otherwise the same unencrypted device. The reason for that is that
467 encryption has an error amplification property: You flip one bit
468 in an encrypted data block, and the decrypted version has half of
469 its bits flipped. This is an important security property for modern
470 ciphers. With the usual modes in cryptsetup (CBC, ESSIV, XTS), you
471 get up to a completely changed 512 byte block per bit error. A
472 corrupt block causes a lot more havoc than the occasionally
473 flipped single bit and can result in various obscure errors.
475 Note, that a verify run on copying between encrypted or
476 unencrypted devices will reliably detect corruption, even when the
477 copying itself did not report any problems. If you find defect
478 RAM, assume all backups and copied data to be suspect, unless you
484 First you should know that overclocking often makes memory
485 problems worse. So if you overclock (which I strongly recommend
486 against in a system holding data that has some worth), run the
487 tests with the overclocking active.
489 There are two good options. One is Memtest86+ and the other is
490 "memtester" by Charles Cazabon. Memtest86+ requires a reboot and
491 then takes over the machine, while memtester runs from a
492 root-shell. Both use different testing methods and I have found
493 problems fast with each one that the other needed long to find. I
494 recommend running the following procedure until the first error is
497 - Run Memtest86+ for one cycle
499 - Run memterster for one cycle (shut down as many other applications
502 - Run Memtest86+ for 24h or more
504 - Run memtester for 24h or more
506 If all that does not produce error messages, your RAM may be sound,
507 but I have had one weak bit that Memtest86+ needed around 60 hours
508 to find. If you can reproduce the original problem reliably, a good
509 additional test may be to remove half of the RAM (if you have more
510 than one module) and try whether the problem is still there and if
511 so, try with the other half. If you just have one module, get a
512 different one and try with that. If you do overclocking, reduce
513 the settings to the most conservative ones available and try with
520 * Is LUKS insecure? Everybody can see I have encrypted data!
522 In practice it does not really matter. In most civilized countries
523 you can just refuse to hand over the keys, no harm done. In some
524 countries they can force you to hand over the keys, if they suspect
525 encryption. However the suspicion is enough, they do not have to
526 prove anything. This is for practical reasons, as even the presence
527 of a header (like the LUKS header) is not enough to prove that you
528 have any keys. It might have been an experiment, for example. Or it
529 was used as encrypted swap with a key from /dev/random. So they
530 make you prove you do not have encrypted data. Of course that is
531 just as impossible as the other way round.
533 This means that if you have a large set of random-looking data,
534 they can already lock you up. Hidden containers (encryption hidden
535 within encryption), as possible with Truecrypt, do not help
536 either. They will just assume the hidden container is there and
537 unless you hand over the key, you will stay locked up. Don't have
538 a hidden container? Though luck. Anybody could claim that.
540 Still, if you are concerned about the LUKS header, use plain
541 dm-crypt with a good passphrase. See also Section 2, "What is the
542 difference between "plain" and LUKS format?"
545 * Should I initialize (overwrite) a new LUKS/dm-crypt partition?
547 If you just create a filesystem on it, most of the old data will
548 still be there. If the old data is sensitive, you should overwrite
549 it before encrypting. In any case, not initializing will leave the
550 old data there until the specific sector gets written. That may
551 enable an attacker to determine how much and where on the
552 partition data was written. If you think this is a risk, you can
553 prevent this by overwriting the encrypted device (here assumed to
554 be named "e1") with zeros like this:
556 dd_rescue -w /dev/zero /dev/mapper/e1
558 or alternatively with one of the following more standard commands:
560 cat /dev/zero > /dev/mapper/e1
561 dd if=/dev/zero of=/dev/mapper/e1
564 * How do I securely erase a LUKS (or other) partition?
566 For LUKS, if you are in a desperate hurry, overwrite the LUKS
567 header and key-slot area. This means overwriting the first
568 (keyslots x stripes x keysize) + offset bytes. For the default
569 parameters, this is the 1'052'672 bytes, i.e. 1MiB + 4096 of the
570 LUKS partition. For 512 bit key length (e.g. for aes-xts-plain with
571 512 bit key) this is 2MiB. (The diferent offset stems from
572 differences in the sector alignment of the key-slots.) If in doubt,
573 just be generous and overwrite the first 10MB or so, it will likely
574 still be fast enough. A single overwrite with zeros should be
575 enough. If you anticipate being in a desperate hurry, prepare the
576 command beforehand. Example with /dev/sde1 as the LUKS partition
577 and default parameters:
579 head -c 1052672 /dev/zero > /dev/sde1; sync
581 A LUKS header backup or full backup will still grant access to
582 most or all data, so make sure that an attacker does not have
583 access to backups or destroy them as well.
585 If you have time, overwrite the whole LUKS partition with a single
586 pass of zeros. This is enough for current HDDs. For SSDs or FLASH
587 (USB sticks) you may want to overwrite the whole drive several
588 times to be sure data is not retained by wear leveling. This is
589 possibly still insecure as SSD technology is not fully understood
590 in this regard. Still, due to the anti-forensic properties of the
591 LUKS key-slots, a single overwrite of an SSD or FLASH drive could
592 be enough. If in doubt, use physical destruction in addition. Here
593 is a link to some current reseach results on erasing SSDs and FLASH
595 http://www.usenix.org/events/fast11/tech/full_papers/Wei.pdf
597 Keep in mind to also erase all backups.
599 Example for a zero-overwrite erase of partition sde1 done with
602 dd_rescue -w /dev/zero /dev/sde1
605 * How do I securely erase a backup of a LUKS partition or header?
607 That depends on the medium it is stored on. For HDD and SSD, use
608 overwrite with zeros. For an SSD or FLASH drive (USB stick), you
609 may want to overwrite the complete SSD several times and use
610 physical destruction in addition, see last item. For re-writable
611 CD/DVD, a single overwrite should also be enough, due to the
612 anti-forensic properties of the LUKS keyslots. For write-once
613 media, use physical destruction. For low security requirements,
614 just cut the CD/DVD into several parts. For high security needs,
615 shred or burn the medium. If your backup is on magnetic tape, I
616 advise physical destruction by shredding or burning, after
617 overwriting . The problem with magnetic tape is that it has a
618 higher dynamic range than HDDs and older data may well be
619 recoverable after overwrites. Also write-head alignment issues can
620 lead to data not actually being deleted at all during overwrites.
623 * What about backup? Does it compromise security?
625 That depends. See next section.
628 * Why is all my data permanently gone if I overwrite the LUKS header?
630 Overwriting the LUKS header in part or in full is the most common
631 reason why access to LUKS containers is lost permanently.
632 Overwriting can be done in a number of fashions, like creating a
633 new filesystem on the raw LUKS partition, making the raw partition
634 part of a raid array and just writing to the raw partition.
636 The LUKS header contains a 256 bit "salt" value and without that no
637 decryption is possible. While the salt is not secret, it is
638 key-grade material and cannot be reconstructed. This is a
639 cryptographically strong "cannot". From observations on the
640 cryptsetup mailing-list, people typically go though the usual
641 stages of grief (Denial, Anger, Bargaining, Depression, Acceptance)
642 when this happens to them. Observed times vary between 1 day and 2
643 weeks to complete the cycle. Seeking help on the mailing-list is
644 fine. Even if we usually cannot help with getting back your data,
645 most people found the feedback comforting.
647 If your header does not contain an intact salt, best go directly
648 to the last stage ("Acceptance") and think about what to do now.
649 There is one exception that I know of: If your LUKS container is
650 still open, then it may be possible to extract the master key from
651 the running system. See Item "How do I recover the master key from
652 a mapped LUKS container?" in Section "Backup and Data Recovery".
657 A salt is a random key-grade value added to the passphrase before
658 it is processed. It is not kept secret. The reason for using salts
659 is as follows: If an attacker wants to crack the password for a
660 single LUKS container, then every possible passphrase has to be
661 tried. Typically an attacker will not try every binary value, but
662 will try words and sentences from a dictionary.
664 If an attacker wants to attack several LUKS containers with the
665 same dictionary, then a different approach makes sense: Compute the
666 resulting slot-key for each dictionary element and store it on
667 disk. Then the test for each entry is just the slow unlocking with
668 the slot key (say 0.00001 sec) instead of calculating the slot-key
669 first (1 sec). For a single attack, this does not help. But if you
670 have more than one container to attack, this helps tremendously,
671 also because you can prepare your table before you even have the
672 container to attack! The calculation is also very simple to
673 parallelize. You could, for example, use the night-time unused CPU
674 power of your desktop PCs for this.
676 This is where the salt comes in. If the salt is combined with the
677 passphrase (in the simplest form, just appended to it), you
678 suddenly need a separate table for each salt value. With a
679 reasonably-sized salt value (256 bit, e.g.) this is quite
683 * Is LUKS secure with a low-entropy (bad) passphrase?
685 Note: You should only use the 94 printable characters from 7 bit
686 ASCII code to prevent your passphrase from failing when the
687 character encoding changes, e.g. because of a system upgrade, see
688 also the note at the very start of this FAQ under "WARNINGS".
690 This needs a bit of theory. The quality of your passphrase is
691 directly related to its entropy (information theoretic, not
692 thermodynamic). The entropy says how many bits of "uncertainty" or
693 "randomness" are in you passphrase. In other words, that is how
694 difficult guessing the passphrase is.
696 Example: A random English sentence has about 1 bit of entropy per
697 character. A random lowercase (or uppercase) character has about
700 Now, if n is the number of bits of entropy in your passphrase and t
701 is the time it takes to process a passphrase in order to open the
702 LUKS container, then an attacker has to spend at maximum
704 attack_time_max = 2^n * t
706 time for a successful attack and on average half that. There is no
707 way getting around that relationship. However, there is one thing
708 that does help, namely increasing t, the time it takes to use a
709 passphrase, see next FAQ item.
711 Still, if you want good security, a high-entropy passphrase is the
712 only option. Use at least 64 bits for secret stuff. That is 64
713 characters of English text (but only if randomly chosen) or a
714 combination of 12 truly random letters and digits.
716 For passphrase generation, do not use lines from very well-known
717 texts (religious texts, Harry potter, etc.) as they are to easy to
718 guess. For example, the total Harry Potter has about 1'500'000
719 words (my estimation). Trying every 64 character sequence starting
720 and ending at a word boundary would take only something like 20
721 days on a single CPU and is entirely feasible. To put that into
722 perspective, using a number of Amazon EC2 High-CPU Extra Large
723 instances (each gives about 8 real cores), this tests costs
724 currently about 50USD/EUR, but can be made to run arbitrarily fast.
726 On the other hand, choosing 1.5 lines from, say, the Wheel of Time
727 is in itself not more secure, but the book selection adds quite a
728 bit of entropy. (Now that I have mentioned it here, don't use tWoT
729 either!) If you add 2 or 3 typos or switch some words around, then
730 this is good passphrase material.
733 * What is "iteration count" and why is decreasing it a bad idea?
735 Iteration count is the number of PBKDF2 iterations a passphrase is
736 put through before it is used to unlock a key-slot. Iterations are
737 done with the explicit purpose to increase the time that it takes
738 to unlock a key-slot. This provides some protection against use of
739 low-entropy passphrases.
741 The idea is that an attacker has to try all possible passphrases.
742 Even if the attacker knows the passphrase is low-entropy (see last
743 item), it is possible to make each individual try take longer. The
744 way to do this is to repeatedly hash the passphrase for a certain
745 time. The attacker then has to spend the same time (given the same
746 computing power) as the user per try. With LUKS, the default is 1
747 second of PBKDF2 hashing.
749 Example 1: Lets assume we have a really bad passphrase (e.g. a
750 girlfriends name) with 10 bits of entropy. With the same CPU, an
751 attacker would need to spend around 500 seconds on average to
752 break that passphrase. Without iteration, it would be more like
753 0.0001 seconds on a modern CPU.
755 Example 2: The user did a bit better and has 32 chars of English
756 text. That would be about 32 bits of entropy. With 1 second
757 iteration, that means an attacker on the same CPU needs around 136
758 years. That is pretty impressive for such a weak passphrase.
759 Without the iterations, it would be more like 50 days on a modern
760 CPU, and possibly far less.
762 In addition, the attacker can both parallelize and use special
763 hardware like GPUs to speed up the attack. The attack can also
764 happen quite some time after the luksFormat operation and CPUs can
765 have become faster and cheaper. For that reason you want a bit
766 of extra security. Anyways, in Example 1 your are screwed. In
767 example 2, not necessarily. Even if the attack is faster, it still
768 has a certain cost associated with it, say 10000 EUR/USD with
769 iteration and 1 EUR/USD without iteration. The first can be
770 prohibitively expensive, while the second is something you try
771 even without solid proof that the decryption will yield something
774 The numbers above are mostly made up, but show the idea. Of course
775 the best thing is to have a high-entropy passphrase.
777 Would a 100 sec iteration time be even better? Yes and no.
778 Cryptographically it would be a lot better, namely 100 times better.
779 However, usability is a very important factor for security
780 technology and one that gets overlooked surprisingly often. For
781 LUKS, if you have to wait 2 minutes to unlock the LUKS container,
782 most people will not bother and use less secure storage instead. It
783 is better to have less protection against low-entropy passphrases
784 and people actually use LUKS, than having them do without
785 encryption altogether.
787 Now, what about decreasing the iteration time? This is generally a
788 very bad idea, unless you know and can enforce that the users only
789 use high-entropy passphrases. If you decrease the iteration time
790 without ensuring that, then you put your users at increased risk,
791 and considering how rarely LUKS containers are unlocked in a
792 typical work-flow, you do so without a good reason. Don't do it.
793 The iteration time is already low enough that users with entropy
794 low passphrases are vulnerable. Lowering it even further increases
795 this danger significantly.
798 * What about iteration count with plain dm-crypt?
800 Simple: There is none. There is also no salting. If you use plain
801 dm-crypt, the only way to be secure is to use a high entropy
802 passphrase. If in doubt, use LUKS instead.
805 * Is LUKS with default parameters less secure on a slow CPU?
807 Unfortunately, yes. However the only aspect affected is the
808 protection for low-entropy passphrase or master-key. All other
809 security aspects are independent of CPU speed.
811 The master key is less critical, as you really have to work at it
812 to give it low entropy. One possibility is to supply the master key
813 yourself. If that key is low-entropy, then you get what you
814 deserve. The other known possibility is to use /dev/urandom for
815 key generation in an entropy-startved situation (e.g. automatic
816 installation on an embedded device without network and other entropy
819 For the passphrase, don't use a low-entropy passphrase. If your
820 passphrase is good, then a slow CPU will not matter. If you insist
821 on a low-entropy passphrase on a slow CPU, use something like
822 "--iter-time=10" or higher and wait a long time on each LUKS unlock
823 and pray that the attacker does not find out in which way exactly
824 your passphrase is low entropy. This also applies to low-entropy
825 passphrases on fast CPUs. Technology can do only so much to
826 compensate for problems in front of the keyboard.
829 * Why was the default aes-cbc-plain replaced with aes-cbc-essiv?
831 The problem is that cbc-plain has a fingerprint vulnerability, where
832 a specially crafted file placed into the crypto-container can be
833 recognized from the outside. The issue here is that for cbc-plain
834 the initialization vector (IV) is the sector number. The IV gets
835 XORed to the first data chunk of the sector to be encrypted. If you
836 make sure that the first data block to be stored in a sector
837 contains the sector number as well, the first data block to be
838 encrypted is all zeros and always encrypted to the same ciphertext.
839 This also works if the first data chunk just has a constant XOR
840 with the sector number. By having several shifted patterns you can
841 take care of the case of a non-power-of-two start sector number of
844 This mechanism allows you to create a pattern of sectors that have
845 the same first ciphertext block and signal one bit per sector to the
846 outside, allowing you to e.g. mark media files that way for
847 recognition without decryption. For large files this is a
848 practical attack. For small ones, you do not have enough blocks to
849 signal and take care of different file starting offsets.
851 In order to prevent this attack, the default was changed to
852 cbc-essiv. ESSIV uses a keyed hash of the sector number, with the
853 encryption key as key. This makes the IV unpredictable without
854 knowing the encryption key and the watermarking attack fails.
857 * Are there any problems with "plain" IV? What is "plain64"?
859 First, "plain" and "plain64" are both not secure to use with CBC,
860 see previous FAQ item.
862 However there are modes, like XTS, that are secure with "plain" IV.
863 The next limit is that "plain" is 64 bit, with the upper 32 bit set
864 to zero. This means that on volumes larger than 2TiB, the IV
865 repeats, creating a vulnerability that potentially leaks some
866 data. To avoid this, use "plain64", which uses the full sector
867 number up to 64 bit. Note that "plain64" requires a kernel >=
868 2.6.33. Also note that "plain64" is backwards compatible for
869 volume sizes <= 2TiB, but not for those > 2TiB. Finally, "plain64"
870 does not cause any performance penalty compared to "plain".
873 * What about XTS mode?
875 XTS mode is potentially even more secure than cbc-essiv (but only if
876 cbc-essiv is insecure in your scenario). It is a NIST standard and
877 used, e.g. in Truecrypt. At the moment, if you want to use it, you
878 have to specify it manually as "aes-xts-plain", i.e.
880 cryptsetup -c aes-xts-plain luksFormat <device>
882 For volumes >2TiB and kernels >= 2.6.33 use "plain64" (see FAQ
883 item on "plain" and "plain64"):
885 cryptsetup -c aes-xts-plain64 luksFormat <device>
887 There is a potential security issue with XTS mode and large blocks.
888 LUKS and dm-crypt always use 512B blocks and the issue does not
892 6. Backup and Data Recovery
895 * Why do I need Backup?
897 First, disks die. The rate for well-treated (!) disk is about 5%
898 per year, which is high enough to worry about. There is some
899 indication that this may be even worse for some SSDs. This applies
900 both to LUKS and plain dm-crypt partitions.
902 Second, for LUKS, if anything damages the LUKS header or the
903 key-stripe area then decrypting the LUKS device can become
904 impossible. This is a frequent occuurence. For example an
905 accidental format as FAT or some software overwriting the first
906 sector where it suspects a partition boot sector typically makes a
907 LUKS partition permanently inacessible. See more below on LUKS
910 So, data-backup in some form is non-optional. For LUKS, you may
911 also want to store a header backup in some secure location. This
912 only needs an update if you change passphrases.
915 * How do I backup a LUKS header?
917 While you could just copy the appropriate number of bytes from the
918 start of the LUKS partition, the best way is to use command option
919 "luksHeaderBackup" of cryptsetup. This protects also against
920 errors when non-standard parameters have been used in LUKS
921 partition creation. Example:
924 cryptsetup luksHeaderBackup --header-backup-file h /dev/mapper/c1
926 To restore, use the inverse command, i.e.
928 cryptsetup luksHeaderRestore --header-backup-file h /dev/mapper/c1
931 * How do I backup a LUKS or dm-crypt partition?
933 There are two options, a sector-image and a plain file or
934 filesystem backup of the contents of the partition. The sector
935 image is already encrypted, but cannot be compressed and contains
936 all empty space. The filesystem backup can be compressed, can
937 contain only part of the encrypted device, but needs to be
938 encrypted separately if so desired.
940 A sector-image will contain the whole partition in encrypted form,
941 for LUKS the LUKS header, the keys-slots and the data area. It can
942 be done under Linux e.g. with dd_rescue (for a direct image copy)
943 and with "cat" or "dd". Example:
945 cat /dev/sda10 > sda10.img
946 dd_rescue /dev/sda10 sda10.img
948 You can also use any other backup software that is capable of making
949 a sector image of a partition. Note that compression is
950 ineffective for encrypted data, hence it does not make sense to
953 For a filesystem backup, you decrypt and mount the encrypted
954 partition and back it up as you would a normal filesystem. In this
955 case the backup is not encrypted, unless your encryption method
956 does that. For example you can encrypt a backup with "tar" as
959 tar cjf - <path> | gpg --cipher-algo AES -c - > backup.tbz2.gpg
961 And verify the backup like this if you are at "path":
963 cat backup.tbz2.gpg | gpg - | tar djf -
965 Note: Allways verify backups, especially encrypted ones.
967 In both cases GnuPG will ask you interactively for your symmetric
968 key. The verify will only output errors. Use "tar dvjf -" to get
969 all comparison results. To make sure no data is written to disk
970 unencrypted, turn off swap if it is not encrypted before doing the
973 You can of course use different or no compression and you can use
974 an asymmetric key if you have one and have a backup of the secret
975 key that belongs to it.
977 A second option for a filestem-level backup that can be used when
978 the backup is also on local disk (e.g. an external USB drive) is
979 to use a LUKS container there and copy the files to be backed up
980 between both mounted containers. Also see next item.
983 * Do I need a backup of the full partition? Would the header and
984 key-slots not be enough?
986 Backup protects you against two things: Disk loss or corruption
987 and user error. By far the most questions on the dm-crypt mailing
988 list about how to recover a damaged LUKS partition are related
989 to user error. For example, if you create a new filesystem on a
990 LUKS partition, chances are good that all data is lost
993 For this case, a header+key-slot backup would often be enough. But
994 keep in mind that a well-treated (!) HDD has roughly a failure
995 risk of 5% per year. It is highly advisable to have a complete
996 backup to protect against this case.
999 * *What do I need to backup if I use "decrypt_derived"?
1001 This is a script in Debian, intended for mounting /tmp or swap with
1002 a key derived from the master key of an already decrypted device.
1003 If you use this for an device with data that should be persistent,
1004 you need to make sure you either do not lose access to that master
1005 key or have a backup of the data. If you derive from a LUKS
1006 device, a header backup of that device would cover backing up the
1007 master key. Keep in mind that this does not protect against disk
1010 Note: If you recreate the LUKS header of the device you derive from
1011 (using luksFormat), the master key changes even if you use the same
1012 passphrase(s) and you will not be able to decrypt the derived
1013 device with the new LUKS header.
1016 * Does a backup compromise security?
1018 Depends on how you do it. However if you do not have one, you are
1019 going to eventually lose your encrypted data.
1021 There are risks introduced by backups. For example if you
1022 change/disable a key-slot in LUKS, a binary backup of the partition
1023 will still have the old key-slot. To deal with this, you have to
1024 be able to change the key-slot on the backup as well, securely
1025 erase the backup or do a filesystem-level backup instead of a binary
1028 If you use dm-crypt, backup is simpler: As there is no key
1029 management, the main risk is that you cannot wipe the backup when
1030 wiping the original. However wiping the original for dm-crypt
1031 should consist of forgetting the passphrase and that you can do
1032 without actual access to the backup.
1034 In both cases, there is an additional (usually small) risk with
1035 binary backups: An attacker can see how many sectors and which
1036 ones have been changed since the backup. To prevent this, use a
1037 filesystem level backup methid that encrypts the whole backup in
1038 one go, e.g. as described above with tar and GnuPG.
1040 My personal advice is to use one USB disk (low value data) or
1041 three disks (high value data) in rotating order for backups, and
1042 either use independent LUKS partitions on them, or use encrypted
1043 backup with tar and GnuPG.
1045 If you do network-backup or tape-backup, I strongly recommend to
1046 go the filesystem backup path with independent encryption, as you
1047 typically cannot reliably delete data in these scenarios,
1048 especially in a cloud setting. (Well, you can burn the tape if it
1049 is under your control...)
1052 * What happens if I overwrite the start of a LUKS partition or damage
1053 the LUKS header or key-slots?
1055 There are two critical components for decryption: The salt values
1056 in the header itself and the key-slots. If the salt values are
1057 overwritten or changed, nothing (in the cryptographically strong
1058 sense) can be done to access the data, unless there is a backup
1059 of the LUKS header. If a key-slot is damaged, the data can still
1060 be read with a different key-slot, if there is a remaining
1061 undamaged and used key-slot. Note that in order to make a key-slot
1062 unrecoverable in a cryptographically strong sense, changing about
1063 4-6 bits in random locations of its 128kiB size is quite enough.
1066 * What happens if I (quick) format a LUKS partition?
1068 I have not tried the different ways to do this, but very likely you
1069 will have written a new boot-sector, which in turn overwrites the
1070 LUKS header, including the salts, making your data permanently
1071 irretrivable, unless you have a LUKS header backup. You may also
1072 damage the key-slots in part or in full. See also last item.
1075 * How do I recover the master key from a mapped LUKS container?
1077 This is typically only needed if you managed to damage your LUKS
1078 header, but the container is still mapped, i.e. "luksOpen"ed.
1080 WARNING: This exposes the master key of the LUKS container. Note
1081 that both ways to recreate a LUKS header with the old master key
1082 described below will write the master key to disk. Unless you are
1083 sure you have securely erased it afterwards, e.g. by writing it to
1084 an encrypted partition, RAM disk or by erasing the filesystem you
1085 wrote it to by a complete overwrite, you should change the master
1086 key afterwards. Changing the master key requires a full data
1087 backup, luksFormat and then restore of the backup.
1089 First, there is a script by Milan that tries to automatize the
1090 whole process, including generating a new LUKS header with the old
1093 http://code.google.com/p/cryptsetup/source/browse/trunk/misc/luks-header-from-active
1095 You can also do this manually. Here is how:
1097 - Get the master key from the device mapper. This is done by the
1098 following command. Substitute c5 for whatever you mapped to:
1100 # dmsetup table --target crypt --showkey /dev/mapper/c5
1102 0 200704 crypt aes-cbc-essiv:sha256
1103 a1704d9715f73a1bb4db581dcacadaf405e700d591e93e2eaade13ba653d0d09
1106 The result is actually one line, wrapped here for clarity. The long
1107 hex string is the master key.
1109 - Convert the master key to a binary file representation. You can
1110 do this manually, e.g. with hexedit. You can also use the tool
1111 "xxd" from vim like this:
1113 echo "a1704d9....53d0d09" | xxd -r -p > master_key
1115 - Do a luksFormat to create a new LUKS header. Unmapthe device
1116 before you do that (luksClose). Replace \dev\dsa10 with the device
1117 the LUKS container is on:
1119 cryptsetup luksFormat --master-key-file=master_key \dev\sda10
1121 Note that if the container was created with other than the default
1122 settings of the cryptsetup version you are using, you need to give
1123 additional parameters specifying the deviations. If in doubt, just
1124 do the first step, keep the whole result safe and try with the
1125 script by Milan. It does recover the other parameters as well.
1127 Side note: This is the way the decrypt_derived script gets at the
1128 master key. It just omits the conversion and hashes the master key
1132 * What does the on-disk structure of dm-crypt look like?
1134 There is none. dm-crypt takes a block device and gives encrypted
1135 access to each of its blocks with a key derived from the passphrase
1136 given. If you use a cipher different than the default, you have to
1137 specify that as a parameter to cryptsetup too. If you want to
1138 change the password, you basically have to create a second
1139 encrypted device with the new passphrase and copy your data over.
1140 On the plus side, if you accidentally overwrite any part of a
1141 dm-crypt device, the damage will be limited to the are you
1145 * What does the on-disk structure of LUKS look like?
1147 A LUKS partition consists of a header, followed by 8 key-slot
1148 descriptors, followed by 8 key slots, followed by the encrypted
1151 Header and key-slot descriptors fill the first 592 bytes. The
1152 key-slot size depends on the creation parameters, namely on the
1153 number of anti-forensic stripes, key material offset and master
1156 With the default parameters, each key-slot is a bit less than
1157 128kiB in size. Due to sector alignment of the key-slot start,
1158 that means the key block 0 is at offset 0x1000-0x20400, key
1159 block 1 at offset 0x21000-0x40400, and key block 7 at offset
1160 0xc1000-0xe0400. The space to the next full sector address is
1161 padded with zeros. Never used key-slots are filled with what the
1162 disk originally contained there, a key-slot removed with
1163 "luksRemoveKey" or "luksKillSlot" gets filled with 0xff. Start of
1164 bulk data is at 0x101000, i.e. at 1'052'672 bytes, i.e. at 1MiB
1165 + 4096 bytes from the start of the partition. This is also the
1166 value given by command "luksDump" with "Payload offset: 2056",
1167 just multiply by the sector size (512 bytes). Incidentally,
1168 "luksHeaderBackup" for a LUKS container created with default
1169 parameters dumps exactly the first 1'052'672 bytes to file and
1170 "luksHeaderRestore" restores them.
1172 For non-default parameters, you have to figure out placement
1173 yourself. "luksDump" helps. For the most common non-default
1174 settings, namely aes-xts-plain with 512 bit key, the offsets are:
1175 1st keyslot 0x1000-0x3f800, 2nd keyslot 0x40000-0x7e000, 3rd
1176 keyslot 0x7e000-0xbd800, ..., and start of bulk data at 0x200000.
1178 The exact specification of the format is here:
1179 http://code.google.com/p/cryptsetup/wiki/Specification
1182 * I think this is overly complicated. Is there an alternative?
1184 Not really. Encryption comes at a price. You can use plain
1185 dm-crypt to simplify things a bit. It does not allow multiple
1186 passphrases, but on the plus side, it has zero on disk description
1187 and if you overwrite some part of a plain dm-crypt partition,
1188 exactly the overwritten parts are lost (rounded up to sector
1192 7. Interoperability with other Disk Encryption Tools
1195 * What is this section about?
1197 Cryptsetup for plain dm-crypt can be used to access a number of
1198 on-disk formats created by tools like loop-aes patched into
1199 losetup. This somtimes works and sometimes does not. This section
1200 collects insights into what works, what does not and where more
1201 information is required.
1203 Additional information may be found in the mailing-list archives,
1204 mentioned at the start of this FAQ document. If you have a
1205 solution working that is not yet documented here and think a wider
1206 audience may be intertested, please email the FAQ maintainer.
1209 * loop-aes: General observations.
1211 One problem is that there are different versions of losetup around.
1212 loop-aes is a patch for losetup. Possible problems and deviations
1213 from cryptsetup option syntax include:
1215 - Offsets specifed in bytes (cryptsetup: 512 byte sectors)
1217 - The need to specify an IV offset
1219 - Encryption mode needs specifying (e.g. "-c twofish-cbc-plain")
1221 - Key size needs specifying (e.g. "-s 128" for 128 bit keys)
1223 - Passphrase hash algorithm needs specifying
1225 Also note that because plain dm-crypt and loop-aes format does not
1226 have metadata, autodetection, while feasible in most cases, would
1227 be a lot of work that nobody really wants to do. If you still have
1228 the old set-up, using a verbosity option (-v) on mapping with the
1229 old tool or having a look into the system logs after setup could
1230 give you the information you need.
1233 * loop-aes patched into losetup on debian 5.x, kernel 2.6.32
1235 In this case, the main problem seems to be that this variant of
1236 losetup takes the offset (-o option) in bytes, while cryptsetup
1237 takes it in sectors of 512 bytes each. Example: The losetupp
1240 losetup -e twofish -o 2560 /dev/loop0 /dev/sdb1
1241 mount /dev/loop0 mountpoint
1245 cryptsetup create -c twofish -o 5 --skip 5 e1 /dev/sdb1
1246 mount /dev/mapper/e1 mountpoint
1249 * loop-aes with 160 bit key
1251 This seems to be sometimes used with twofish and blowfish and
1252 represents a 160 bit ripemed160 hash output padded to 196 bit key
1253 length. It seems the corresponding options for cryptsetup are
1255 --cipher twofish-cbc-null -s 192 -h ripemd160:20
1258 8. Issues with Specific Versions of cryptsetup
1261 * When using the create command for plain dm-crypt with cryptsetup
1262 1.1.x, the mapping is incompatible and my data is not accessible
1265 With cryptsetup 1.1.x, the distro maintainer can define different
1266 default encryption modes for LUKS and plain devices. You can check
1267 these compiled-in defaults using "cryptsetup --help". Moreover, the
1268 plain device default changed because the old IV mode was
1269 vulnerable to a watermarking attack.
1271 If you are using a plain device and you need a compatible mode, just
1272 specify cipher, key size and hash algorithm explicitly. For
1273 compatibility with cryptsetup 1.0.x defaults, simple use the
1276 cryptsetup create -c aes-cbc-plain -s 256 -h ripemd160 <name> <dev>
1278 LUKS stores cipher and mode in the metadata on disk, avoiding this
1282 * cryptsetup on SLED 10 has problems...
1284 SLED 10 is missing an essential kernel patch for dm-crypt, which
1285 is broken in its kernel as a result. There may be a very old
1286 version of cryptsetup (1.0.x) provided by SLED, which should also
1287 not be used anymore as well. My advice would be to drop SLED 10.
1289 A. Contributors In no particular order: