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 just managed to
33 somehow format or overwrite the start of their LUKS partitions. 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 such
39 PASSPHRASES: Some people have had difficulties when upgrading
40 distributions. It is highly advisable to only use the 94 printable
41 characters from the first 128 characters of the ASCII table, as
42 they will always have the same binary representation. Other
43 characters may have different encoding depending on system
44 configuration and your passphrase will not work with a different
45 encoding. A table of the standardized first 128 ASCII caracters
46 can, e.g. be found on http://en.wikipedia.org/wiki/ASCII
49 * System Specific warnings
51 - Ubuntu as of 4/2011: It seems the installer offers to create
52 LUKS partitions in a way that several people mistook for an offer
53 to activate their existing LUKS partition. The installer gives no
54 or an inadequate warning and will destroy your old LUKS header,
55 causing permanent data loss. See also the section on Backup and
58 This issue has been acknowledged by the Ubuntu dev team, see here:
59 http://launchpad.net/bugs/420080
64 Current FAQ maintainer is Arno Wagner <arno@wagner.name>. Other
65 contributors are listed at the end. If you want to contribute, send
66 your article, including a descriptive headline, to the maintainer,
67 or the dm-crypt mailing list with something like "FAQ ..." in the
68 subject. You can also send more raw information and have me write
69 the section. Please note that by contributing to this FAQ, you
70 accept the license described below.
72 This work is under the "Attribution-Share Alike 3.0 Unported"
73 license, which means distribution is unlimited, you may create
74 derived works, but attributions to original authors and this
75 license statement must be retained and the derived work must be
76 under the same license. See
77 http://creativecommons.org/licenses/by-sa/3.0/ for more details of
80 Side note: I did text license research some time ago and I think
81 this license is best suited for the purpose at hand and creates the
85 * Where is the project website?
87 There is the project website at http://code.google.com/p/cryptsetup/
88 Please do not post questions there, nobody will read them. Use
89 the mailing-list instead.
92 * Is there a mailing-list?
94 Instructions on how to subscribe to the mailing-list are at on the
95 project website. People are generally helpful and friendly on the
98 The question of how to unsubscribe from the list does crop up
99 sometimes. For this you need your list management URL, which is
100 sent to you initially and once at the start of each month. Go to
101 the URL mentioned in the email and select "unsubscribe". This page
102 also allows you to request a password reminder.
104 Alternatively, you can send an Email to dm-crypt-request@saout.de
105 with just the word "help" in the subject or message body. Make sure
106 to send it from your list address.
108 The mailing list archive is here:
109 http://dir.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt
115 * What is the difference between "plain" and LUKS format?
117 Plain format is just that: It has no metadata on disk, reads all
118 paramters from the commandline (or the defaults), derives a
119 master-key from the passphrase and then uses that to de-/encrypt
120 the sectors of the device, with a direct 1:1 mapping between
121 encrypted and decrypted sectors.
123 Primary advantage is high resilience to damage, as one damaged
124 encrypted sector results in exactly one damaged decrypted sector.
125 Also, it is not readily apparent that there even is encrypted data
126 on the device, as an overwrite with crypto-grade randomness (e.g.
127 from /dev/urandom) looks exactly the same on disk.
129 Side-note: That has limited value against the authorities. In
130 civilized countries, they cannot force you to give up a crypto-key
131 anyways. In the US, the UK and dictatorships around the world,
132 they can force you to give up the keys (using imprisonment or worse
133 to pressure you), and in the worst case, they only need a
134 nebulous "suspicion" about the presence of encrypted data. My
135 advice is to either be ready to give up the keys or to not have
136 encrypted data when traveling to those countries, especially when
137 crossing the borders.
139 Disadvantages are that you do not have all the nice features that
140 the LUKS metadata offers, like multiple passphrases that can be
141 changed, the cipher being stored in the metadata, anti-forensic
142 properties like key-slot diffusion and salts, etc..
144 LUKS format uses a metadata header and 8 key-slot areas that are
145 being placed ath the begining of the disk, see below under "What
146 does the LUKS on-disk format looks like?". The passphrases are used
147 to decryt a single master key that is stored in the anti-forensic
150 Advantages are a higher usability, automatic configuration of
151 non-default crypto parameters, defenses against low-entropy
152 passphrases like salting and iterated PBKDF2 passphrase hashing,
153 the ability to change passhrases, and others.
155 Disadvantages are that it is readily obvious there is encrypted
156 data on disk (but see side note above) and that damage to the
157 header or key-slots usually results in permanent data-loss. See
158 below under "6. Backup and Data Recovery" on how to reduce that
159 risk. Also the sector numbers get shifted by the length of the
160 header and key-slots and there is a loss of that size in capacity
161 (1MB+4096B for defaults and 2MB for the most commonly used
162 non-default XTS mode).
165 * Can I encrypt an already existing, non-empty partition to use LUKS?
167 There is no converter, and it is not really needed. The way to do
168 this is to make a backup of the device in question, securely wipe
169 the device (as LUKS device initialization does not clear away old
170 data), do a luksFormat, optionally overwrite the encrypted device,
171 create a new filesystem and restore your backup on the now
172 encrypted device. Also refer to sections "Security Aspects" and
173 "Backup and Data Recovery".
175 For backup, plain GNU tar works well and backs up anything likely
176 to be in a filesystem.
179 * How do I use LUKS with a loop-device?
181 This can be very handy for experiments. Setup is just the same as
182 with any block device. If you want, for example, to use a 100MiB
183 file as LUKS container, do something like this:
185 head -c 100M /dev/zero > luksfile # create empty file
186 losetup /dev/loop0 luksfile # map luksfile to /dev/loop0
187 cryptsetup luksFormat /dev/loop0 # create LUKS on loop device
189 Afterwards just use /dev/loop0 as a you would use a LUKS partition.
190 To unmap the file when done, use "losetup -d /dev/loop0".
193 * When I add a new key-slot to LUKS, it asks for a passphrase but
194 then complains about there not being a key-slot with that
197 That is as intended. You are asked a passphrase of an existing
198 key-slot first, before you can enter the passphrase for the new
199 key-slot. Otherwise you could break the encryption by just adding a
200 new key-slot. This way, you have to know the passphrase of one of
201 the already configured key-slots in order to be able to configure a
205 * How do I read a dm-crypt key from file?
207 Note that the file will still be hashed first, just like keyboard
208 input. Use the --key-file option, like this:
210 cryptsetup create --key-file keyfile e1 /dev/loop0
213 * How do I read a LUKS slot key from file?
215 What you really do here is to read a passphrase from file, just as
216 you would with manual entry of a passphrase for a key-slot. You can
217 add a new passphrase to a free key-slot, set the passphrase of an
218 specific key-slot or put an already configured passphrase into a
219 file. In the last case make sure no trailing newline (0x0a) is
220 contained in the key file, or the passphrase will not work because
221 the whole file is used as input.
223 To add a new passphrase to a free key slot from file, use something
226 cryptsetup luksAddKey /dev/loop0 keyfile
228 To add a new passphrase to a specific key-slot, use something like
231 cryptsetup luksAddKey --key-slot 7 /dev/loop0 keyfile
233 To supply a key from file to any LUKS command, use the --key-file
234 option, e.g. like this:
236 cryptsetup luksOpen --key-file keyfile /dev/loop0 e1
239 * How do I read the LUKS master key from file?
241 The question you should ask yourself first is why you would want to
242 do this. The only legitimate reason I can think of is if you want
243 to have two LUKS devices with the same master key. Even then, I
244 think it would be preferable to just use key-slots with the same
245 passphrase, or to use plain dm-crypt instead. If you really have a
246 good reason, please tell me. If I am convinced, I will add how to
250 * What are the security requirements for a key read from file?
252 A file-stored key or passphrase has the same security requirements
253 as one entered interactively, however you can use random bytes and
254 thereby use bytes you cannot type on the keyboard. You can use any
255 file you like as key file, for example a plain text file with a
256 human readable passphrase. To generate a file with random bytes,
257 use something like this:
259 head -c 256 /dev/random > keyfile
262 * If I map a journaled file system using dm-crypt/LUKS, does it still
263 provide its usual transactional guarantees?
265 As far as I know it does (but I may be wrong), but please note that
266 these "guarantees" are far weaker than they appear to be. For
267 example, you may not get a hard flush to disk surface even on a
268 call to fsync. In addition, the HDD itself may do independent
269 write reordering. Some other things can go wrong as well. The
270 filesystem developers are aware of these problems and typically
271 can make it work anyways. That said, dm-crypt/LUKS should not make
274 Personally, I have several instances of ext3 on dm-crypt and have
275 not noticed any specific problems.
277 Update: I did run into frequent small freezes (1-2 sec) when putting
278 a vmware image on ext3 over dm-crypt. This does indicate that the
279 transactional guarantees are in place, but at a cost. When I went
280 back to ext2, the problem went away. This also seems to have gotten
281 better with kernel 2.6.36 and the reworking of filesystem flush
282 locking. Kernel 2.6.38 is expected to have more improvements here.
285 * Can I use LUKS or cryptsetup with a more secure (external) medium
286 for key storage, e.g. TPM or a smartcard?
288 Yes, see the answers on using a file-supplied key. You do have to
289 write the glue-logic yourself though. Basically you can have
290 cryptsetup read the key from STDIN and write it there with your
291 own tool that in turn gets the key from the more secure key
295 * Can I resize a dm-crypt or LUKS partition?
297 Yes, you can, as neither dm-crypt nor LUKS stores partition size.
298 Whether you should is a different question. Personally I recommend
299 backup, recreation of the encrypted partition with new size,
300 recreation of the filesystem and restore. This gets around the
301 tricky business of resizing the filesystem. Resizing a dm-crypt or
302 LUKS container does not resize the filesystem in it. The backup is
303 really non-optional here, as a lot can go wrong, resulting in
304 partial or complete data loss. Using something like gparted to
305 resize an encrypted partition is slow, but typicaly works. This
306 will not change the size of the filesystem hidden under the
309 You also need to be aware of size-based limitations. The one
310 currently relevant is that aes-xts-plain should not be used for
311 encrypted container sizes larger than 2TiB. Use aes-xts-plain64
318 * My dm-crypt/LUKS mapping does not work! What general steps are
319 there to investigate the problem?
321 If you get a specific error message, investigate what it claims
322 first. If not, you may want to check the following things.
324 - Check that "/dev", including "/dev/mapper/control" is there. If it
325 is missing, you may have a problem with the "/dev" tree itself or
326 you may have broken udev rules.
328 - Check that you have the device mapper and the crypt target in your
329 kernel. The output of "dmsetup targets" should list a "crypt"
330 target. If it is not there or the command fails, add device mapper
331 and crypt-target to the kernel.
333 - Check that the hash-functions and ciphers you want to use are in
334 the kernel. The output of "cat /proc/crypto" needs to list them.
337 * My dm-crypt mapping suddenly stopped when upgrading cryptsetup.
339 The default cipher, hash or mode may have changed (the mode changed
340 from 1.0.x to 1.1.x). See under "Issues With Specific Versions of
344 * When I call cryptsetup from cron/CGI, I get errors about unknown
347 If you get errors about unknown parameters or the like that are not
348 present when cryptsetup is called from the shell, make sure you
349 have no older version of cryptsetup on your system that then gets
350 called by cron/CGI. For example some distributions install
351 cryptsetup into /usr/sbin, while a manual install could go to
352 /usr/local/sbin. As a debugging aid, call "cryptsetup --version"
353 from cron/CGI or the non-shell mechanism to be sure the right
357 * Unlocking a LUKS device takes very long. Why?
359 The iteration time for a key-slot (see Section 5 for an explanation
360 what iteration does) is calculated when setting a passphrase. By
361 default it is 1 second on the machine where the passphrase is set.
362 If you set a passphrase on a fast machine and then unlock it on a
363 slow machine, the unlocking time can be much longer. Also take into
364 account that up to 8 key-slots have to be tried in order to find the
367 If this is problem, you can add another key-slot using the slow
368 machine with the same passphrase and then remove the old key-slot.
369 The new key-slot will have an iteration count adjusted to 1 second
370 on the slow machine. Use luksKeyAdd and then luksKillSlot or
373 However, this operation will not change volume key iteration count
374 (MK iterations in output of "cryptsetup luksDump"). In order to
375 change that, you will have to backup the data in the LUKS
376 container, luksFormat on the slow machine and restore the data.
377 Note that in the original LUKS specification this value was fixed
378 to 10, but it is now derived from the PBKDF2 benchmark as well and
379 set to iterations in 0.125 sec or 1000, whichever is larger.
382 * "blkid" sees a LUKS UUID and an ext2/swap UUID on the same device.
385 Some old versions of cryptsetup have a bug where the header does
386 not get completely wiped during LUKS format and an older ext2/swap
387 signature remains on the device. This confuses blkid.
389 Fix: Wipe the unused header areas by doing a backup and restore of
390 the header with cryptsetup 1.1.x:
392 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
393 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
396 * cryptsetup segfaults on Gentoo amd64 hardened ...
398 There seems to be some inteference between the hardening and and
399 the way cryptsetup benchmarks PBKDF2. The solution to this is
400 currently not quite clear for an encrypted root filesystem. For
401 other uses, you can apparently specify USE="dynamic" as compile
402 flag, see http://bugs.gentoo.org/show_bug.cgi?id=283470
408 * Can a bad RAM module cause problems?
410 LUKS and dm-crypt can give the RAM quite a workout, especially when
411 combined with software RAID. In particular the combination RAID5 +
412 LUKS + XFS seems to uncover RAM problems that never caused obvious
413 problems before. Symptoms vary, but often the problem manifest
414 itself when copying large amounts of data, typically several times
415 larger than your main memory.
417 Side note: One thing you should always do on large data
418 copy/movements is to run a verify, for example with the "-d"
419 option of "tar" or by doing a set of MD5 checksums on the source
422 find . -type f -exec md5sum \{\} \; > checksum-file
424 and then a "md5sum -c checksum-file" on the other side. If you get
425 mismatches here, RAM is the primary suspect. A lesser suspect is
426 an overclocked CPU. I have found countless hardware problems in
427 verify runs after copying or making backups. Bit errors are much
428 more common than most people think.
430 Some RAM issues are even worse and corrupt structures in one of the
431 layers. This typically results in lockups, CPU state dumps in the
432 system logs, kernel panic or other things. It is quite possible to
433 have the problem with an encrypted device, but not with an
434 otherwise the same unencrypted device. The reason for that is that
435 encryption has an error amplification property: You flip one bit
436 in an encrypted data block, and the decrypted version has half of
437 its bits flipped. This is an important security property for modern
438 ciphers. With the usual modes in cryptsetup (CBC, ESSIV, XTS), you
439 get up to a completely changed 512 byte block per bit error. A
440 corrupt block causes a lot more havoc than the occasionally
441 flipped single bit and can result in various obscure errors.
443 Note, that a verify run on copying between encrypted or
444 unencrypted devices will reliably detect corruption, even when the
445 copying itself did not report any problems. If you find defect
446 RAM, assume all backups and copied data to be suspect, unless you
452 First you should know that overclocking often makes memory
453 problems worse. So if you overclock (which I strongly recommend
454 against in a system holding data that has some worth), run the
455 tests with the overclocking active.
457 There are two good options. One is Memtest86+ and the other is
458 "memtester" by Charles Cazabon. Memtest86+ requires a reboot and
459 then takes over the machine, while memtester runs from a
460 root-shell. Both use different testing methods and I have found
461 problems fast with each one that the other needed long to find. I
462 recommend running the following procedure until the first error is
465 - Run Memtest86+ for one cycle
467 - Run memterster for one cycle (shut down as many other applications
470 - Run Memtest86+ for 24h or more
472 - Run memtester for 24h or more
474 If all that does not produce error messages, your RAM may be sound,
475 but I have had one weak bit that Memtest86+ needed around 60 hours
476 to find. If you can reproduce the original problem reliably, a good
477 additional test may be to remove half of the RAM (if you have more
478 than one module) and try whether the problem is still there and if
479 so, try with the other half. If you just have one module, get a
480 different one and try with that. If you do overclocking, reduce
481 the settings to the most conservative ones available and try with
488 * Is LUKS insecure? Everybody can see I have encrypted data!
490 In practice it does not really matter. In most civilized countries
491 you can just refuse to hand over the keys, no harm done. In some
492 countries they can force you to hand over the keys, if they suspect
493 encryption. However the suspicion is enough, they do not have to
494 prove anything. This is for practical reasons, as even the presence
495 of a header (like the LUKS header) is not enough to prove that you
496 have any keys. It might have been an experiment, for example. Or it
497 was used as encrypted swap with a key from /dev/random. So they
498 make you prove you do not have encrypted data. Of course that is
499 just as impossible as the other way round.
501 This means that if you have a large set of random-looking data,
502 they can already lock you up. Hidden containers (encryption hidden
503 within encryption), as possible with Truecrypt, do not help
504 either. They will just assume the hidden container is there and
505 unless you hand over the key, you will stay locked up. Don't have
506 a hidden container? Though luck. Anybody could claim that.
508 Still, if you are concerned about the LUKS header, use plain
509 dm-crypt with a good passphrase. See also Section 2, "What is the
510 difference between "plain" and LUKS format?"
513 * Should I initialize (overwrite) a new LUKS/dm-crypt partition?
515 If you just create a filesystem on it, most of the old data will
516 still be there. If the old data is sensitive, you should overwrite
517 it before encrypting. In any case, not initializing will leave the
518 old data there until the specific sector gets written. That may
519 enable an attacker to determine how much and where on the
520 partition data was written. If you think this is a risk, you can
521 prevent this by overwriting the encrypted device (here assumed to
522 be named "e1") with zeros like this:
524 dd_rescue -w /dev/zero /dev/mapper/e1
526 or alternatively with one of the following more standard commands:
528 cat /dev/zero > /dev/mapper/e1
529 dd if=/dev/zero of=/dev/mapper/e1
532 * How do I securely erase a LUKS (or other) partition?
534 For LUKS, if you are in a desperate hurry, overwrite the LUKS
535 header and key-slot area. This means overwriting the first
536 (keyslots x stripes x keysize) + offset bytes. For the default
537 parameters, this is the 1'052'672 bytes, i.e. 1MiB + 4096 of the
538 LUKS partition. For 512 bit key length (e.g. for aes-xts-plain with
539 512 bit key) this is 2MiB. (The diferent offset stems from
540 differences in the sector alignment of the key-slots.) If in doubt,
541 just be generous and overwrite the first 10MB or so, it will likely
542 still be fast enough. A single overwrite with zeros should be
543 enough. If you anticipate being in a desperate hurry, prepare the
544 command beforehand. Example with /dev/sde1 as the LUKS partition
545 and default parameters:
547 head -c 1052672 /dev/zero > /dev/sde1; sync
549 A LUKS header backup or full backup will still grant access to
550 most or all data, so make sure that an attacker does not have
551 access to backups or destroy them as well.
553 If you have time, overwrite the whole LUKS partition with a single
554 pass of zeros. This is enough for current HDDs. For SSDs or FLASH
555 (USB sticks) you may want to overwrite the whole drive several
556 times to be sure data is not retained by wear leveling. This is
557 possibly still insecure as SSD technology is not fully understood
558 in this regard. Still, due to the anti-forensic properties of the
559 LUKS key-slots, a single overwrite of an SSD or FLASH drive could
560 be enough. If in doubt, use physical destruction in addition. Here
561 is a link to some current reseach results on erasing SSDs and FLASH
563 http://www.usenix.org/events/fast11/tech/full_papers/Wei.pdf
565 Keep in mind to also erase all backups.
567 Example for a zero-overwrite erase of partition sde1 done with
570 dd_rescue -w /dev/zero /dev/sde1
573 * How do I securely erase a backup of a LUKS partition or header?
575 That depends on the medium it is stored on. For HDD and SSD, use
576 overwrite with zeros. For an SSD or FLASH drive (USB stick), you
577 may want to overwrite the complete SSD several times and use
578 physical destruction in addition, see last item. For re-writable
579 CD/DVD, a single overwrite should also be enough, due to the
580 anti-forensic properties of the LUKS keyslots. For write-once
581 media, use physical destruction. For low security requirements,
582 just cut the CD/DVD into several parts. For high security needs,
583 shred or burn the medium. If your backup is on magnetic tape, I
584 advise physical destruction by shredding or burning, after
585 overwriting . The problem with magnetic tape is that it has a
586 higher dynamic range than HDDs and older data may well be
587 recoverable after overwrites. Also write-head alignment issues can
588 lead to data not actually being deleted at all during overwrites.
591 * What about backup? Does it compromise security?
593 That depends. See next section.
596 * Why is all my data permanently gone if I overwrite the LUKS header?
598 Overwriting the LUKS header in part or in full is the most common
599 reason why access to LUKS containers is lost permanently.
600 Overwriting can be done in a number of fashions, like creating a
601 new filesystem on the raw LUKS partition, making the raw partition
602 part of a raid array and just writing to the raw partition.
604 The LUKS header contains a 256 bit "salt" value and without that no
605 decryption is possible. While the salt is not secret, it is
606 key-grade material and cannot be reconstructed. This is a
607 cryptographically strong "cannot". From observations on the
608 cryptsetup mailing-list, people typically go though the usual
609 stages of grief (Denial, Anger, Bargaining, Depression, Acceptance)
610 when this happens to them. Observed times vary between 1 day and 2
611 weeks to complete the cycle. Seeking help on the mailing-list is
612 fine. Even if we usually cannot help with getting back your data,
613 most people found the feedback comforting.
615 If your header does not contain an intact salt, best go directly
616 to the last stage ("Acceptance") and think about what to do now.
617 There is one exception that I know of: If your LUKS container is
618 still open, then it may be possible to extract the master key from
619 the running system. Ask on the mailing-list on how to do that and
620 make sure nobody switches off the machine.
625 A salt is a random key-grade value added to the passphrase before
626 it is processed. It is not kept secret. The reason for using salts
627 is as follows: If an attacker wants to crack the password for a
628 single LUKS container, then every possible passphrase has to be
629 tried. Typically an attacker will not try every binary value, but
630 will try words and sentences from a dictionary.
632 If an attacker wants to attack several LUKS containers with the
633 same dictionary, then a different approach makes sense: Compute the
634 resulting slot-key for each dictionary element and store it on
635 disk. Then the test for each entry is just the slow unlocking with
636 the slot key (say 0.00001 sec) instead of calculating the slot-key
637 first (1 sec). For a single attack, this does not help. But if you
638 have more than one container to attack, this helps tremendously,
639 also because you can prepare your table before you even have the
640 container to attack! The calculation is also very simple to
641 parallelize. You could, for example, use the night-time unused CPU
642 power of your desktop PCs for this.
644 This is where the salt comes in. If the salt is combined with the
645 passphrase (in the simplest form, just appended to it), you
646 suddenly need a separate table for each salt value. With a
647 reasonably-sized salt value (256 bit, e.g.) this is quite
651 * Is LUKS secure with a low-entropy (bad) passphrase?
653 Note: You should only use the 94 printable characters from 7 bit
654 ASCII code to prevent your passphrase from failing when the
655 character encoding changes, e.g. because of a system upgrade, see
656 also the note at the very start of this FAQ under "WARNINGS".
658 This needs a bit of theory. The quality of your passphrase is
659 directly related to its entropy (information theoretic, not
660 thermodynamic). The entropy says how many bits of "uncertainty" or
661 "randomness" are in you passphrase. In other words, that is how
662 difficult guessing the passphrase is.
664 Example: A random English sentence has about 1 bit of entropy per
665 character. A random lowercase (or uppercase) character has about
668 Now, if n is the number of bits of entropy in your passphrase and t
669 is the time it takes to process a passphrase in order to open the
670 LUKS container, then an attacker has to spend at maximum
672 attack_time_max = 2^n * t
674 time for a successful attack and on average half that. There is no
675 way getting around that relationship. However, there is one thing
676 that does help, namely increasing t, the time it takes to use a
677 passphrase, see next FAQ item.
679 Still, if you want good security, a high-entropy passphrase is the
680 only option. Use at least 64 bits for secret stuff. That is 64
681 characters of English text (but only if randomly chosen) or a
682 combination of 12 truly random letters and digits.
684 For passphrase generation, do not use lines from very well-known
685 texts (religious texts, Harry potter, etc.) as they are to easy to
686 guess. For example, the total Harry Potter has about 1'500'000
687 words (my estimation). Trying every 64 character sequence starting
688 and ending at a word boundary would take only something like 20
689 days on a single CPU and is entirely feasible. To put that into
690 perspective, using a number of Amazon EC2 High-CPU Extra Large
691 instances (each gives about 8 real cores), this tests costs
692 currently about 50USD/EUR, but can be made to run arbitrarily fast.
694 On the other hand, choosing 1.5 lines from, say, the Wheel of Time
695 is in itself not more secure, but the book selection adds quite a
696 bit of entropy. (Now that I have mentioned it here, don't use tWoT
697 either!) If you add 2 or 3 typos or switch some words around, then
698 this is good passphrase material.
701 * What is "iteration count" and why is decreasing it a bad idea?
703 Iteration count is the number of PBKDF2 iterations a passphrase is
704 put through before it is used to unlock a key-slot. Iterations are
705 done with the explicit purpose to increase the time that it takes
706 to unlock a key-slot. This provides some protection against use of
707 low-entropy passphrases.
709 The idea is that an attacker has to try all possible passphrases.
710 Even if the attacker knows the passphrase is low-entropy (see last
711 item), it is possible to make each individual try take longer. The
712 way to do this is to repeatedly hash the passphrase for a certain
713 time. The attacker then has to spend the same time (given the same
714 computing power) as the user per try. With LUKS, the default is 1
715 second of PBKDF2 hashing.
717 Example 1: Lets assume we have a really bad passphrase (e.g. a
718 girlfriends name) with 10 bits of entropy. With the same CPU, an
719 attacker would need to spend around 500 seconds on average to
720 break that passphrase. Without iteration, it would be more like
721 0.0001 seconds on a modern CPU.
723 Example 2: The user did a bit better and has 32 chars of English
724 text. That would be about 32 bits of entropy. With 1 second
725 iteration, that means an attacker on the same CPU needs around 136
726 years. That is pretty impressive for such a weak passphrase.
727 Without the iterations, it would be more like 50 days on a modern
728 CPU, and possibly far less.
730 In addition, the attacker can both parallelize and use special
731 hardware like GPUs to speed up the attack. The attack can also
732 happen quite some time after the luksFormat operation and CPUs can
733 have become faster and cheaper. For that reason you want a bit
734 of extra security. Anyways, in Example 1 your are screwed. In
735 example 2, not necessarily. Even if the attack is faster, it still
736 has a certain cost associated with it, say 10000 EUR/USD with
737 iteration and 1 EUR/USD without iteration. The first can be
738 prohibitively expensive, while the second is something you try
739 even without solid proof that the decryption will yield something
742 The numbers above are mostly made up, but show the idea. Of course
743 the best thing is to have a high-entropy passphrase.
745 Would a 100 sec iteration time be even better? Yes and no.
746 Cryptographically it would be a lot better, namely 100 times better.
747 However, usability is a very important factor for security
748 technology and one that gets overlooked surprisingly often. For
749 LUKS, if you have to wait 2 minutes to unlock the LUKS container,
750 most people will not bother and use less secure storage instead. It
751 is better to have less protection against low-entropy passphrases
752 and people actually use LUKS, than having them do without
753 encryption altogether.
755 Now, what about decreasing the iteration time? This is generally a
756 very bad idea, unless you know and can enforce that the users only
757 use high-entropy passphrases. If you decrease the iteration time
758 without ensuring that, then you put your users at increased risk,
759 and considering how rarely LUKS containers are unlocked in a
760 typical work-flow, you do so without a good reason. Don't do it.
761 The iteration time is already low enough that users with entropy
762 low passphrases are vulnerable. Lowering it even further increases
763 this danger significantly.
766 * What about iteration count with plain dm-crypt?
768 Simple: There is none. There is also no salting. If you use plain
769 dm-crypt, the only way to be secure is to use a high entropy
770 passphrase. If in doubt, use LUKS instead.
773 * Is LUKS with default parameters less secure on a slow CPU?
775 Unfortunately, yes. However the only aspect affected is the
776 protection for low-entropy passphrase or master-key. All other
777 security aspects are independent of CPU speed.
779 The master key is less critical, as you really have to work at it
780 to give it low entropy. One possibility is to supply the master key
781 yourself. If that key is low-entropy, then you get what you
782 deserve. The other known possibility is to use /dev/urandom for
783 key generation in an entropy-startved situation (e.g. automatic
784 installation on an embedded device without network and other entropy
787 For the passphrase, don't use a low-entropy passphrase. If your
788 passphrase is good, then a slow CPU will not matter. If you insist
789 on a low-entropy passphrase on a slow CPU, use something like
790 "--iter-time=10" or higher and wait a long time on each LUKS unlock
791 and pray that the attacker does not find out in which way exactly
792 your passphrase is low entropy. This also applies to low-entropy
793 passphrases on fast CPUs. Technology can do only so much to
794 compensate for problems in front of the keyboard.
797 * Why was the default aes-cbc-plain replaced with aes-cbc-essiv?
799 The problem is that cbc-plain has a fingerprint vulnerability, where
800 a specially crafted file placed into the crypto-container can be
801 recognized from the outside. The issue here is that for cbc-plain
802 the initialization vector (IV) is the sector number. The IV gets
803 XORed to the first data chunk of the sector to be encrypted. If you
804 make sure that the first data block to be stored in a sector
805 contains the sector number as well, the first data block to be
806 encrypted is all zeros and always encrypted to the same ciphertext.
807 This also works if the first data chunk just has a constant XOR
808 with the sector number. By having several shifted patterns you can
809 take care of the case of a non-power-of-two start sector number of
812 This mechanism allows you to create a pattern of sectors that have
813 the same first ciphertext block and signal one bit per sector to the
814 outside, allowing you to e.g. mark media files that way for
815 recognition without decryption. For large files this is a
816 practical attack. For small ones, you do not have enough blocks to
817 signal and take care of different file starting offsets.
819 In order to prevent this attack, the default was changed to
820 cbc-essiv. ESSIV uses a keyed hash of the sector number, with the
821 encryption key as key. This makes the IV unpredictable without
822 knowing the encryption key and the watermarking attack fails.
825 * Are there any problems with "plain" IV? What is "plain64"?
827 First, "plain" and "plain64" are both not secure to use with CBC,
828 see previous FAQ item.
830 However there are modes, like XTS, that are secure with "plain" IV.
831 The next limit is that "plain" is 64 bit, with the upper 32 bit set
832 to zero. This means that on volumes larger than 2TiB, the IV
833 repeats, creating a vulnerability that potentially leaks some
834 data. To avoid this, use "plain64", which uses the full sector
835 number up to 64 bit. Note that "plain64" requires a kernel >=
836 2.6.33. Also note that "plain64" is backwards compatible for
837 volume sizes <= 2TiB, but not for those > 2TiB. Finally, "plain64"
838 does not cause any performance penalty compared to "plain".
841 * What about XTS mode?
843 XTS mode is potentially even more secure than cbc-essiv (but only if
844 cbc-essiv is insecure in your scenario). It is a NIST standard and
845 used, e.g. in Truecrypt. At the moment, if you want to use it, you
846 have to specify it manually as "aes-xts-plain", i.e.
848 cryptsetup -c aes-xts-plain luksFormat <device>
850 For volumes >2TiB and kernels >= 2.6.33 use "plain64" (see FAQ
851 item on "plain" and "plain64"):
853 cryptsetup -c aes-xts-plain64 luksFormat <device>
855 There is a potential security issue with XTS mode and large blocks.
856 LUKS and dm-crypt always use 512B blocks and the issue does not
860 6. Backup and Data Recovery
863 * Why do I need Backup?
865 First, disks die. The rate for well-treated (!) disk is about 5%
866 per year, which is high enough to worry about. There is some
867 indication that this may be even worse for some SSDs. This applies
868 both to LUKS and plain dm-crypt partitions.
870 Second, for LUKS, if anything damages the LUKS header or the
871 key-stripe area then decrypting the LUKS device can become
872 impossible. This is a frequent occuurence. For example an
873 accidental format as FAT or some software overwriting the first
874 sector where it suspects a partition boot sector typically makes a
875 LUKS partition permanently inacessible. See more below on LUKS
878 So, data-backup in some form is non-optional. For LUKS, you may
879 also want to store a header backup in some secure location. This
880 only needs an update if you change passphrases.
883 * How do I backup a LUKS header?
885 While you could just copy the appropriate number of bytes from the
886 start of the LUKS partition, the best way is to use command option
887 "luksHeaderBackup" of cryptsetup. This protects also against
888 errors when non-standard parameters have been used in LUKS
889 partition creation. Example:
892 cryptsetup luksHeaderBackup --header-backup-file h /dev/mapper/c1
894 To restore, use the inverse command, i.e.
896 cryptsetup luksHeaderRestore --header-backup-file h /dev/mapper/c1
899 * How do I backup a LUKS or dm-crypt partition?
901 There are two options, a sector-image and a plain file or
902 filesystem backup of the contents of the partition. The sector
903 image is already encrypted, but cannot be compressed and contains
904 all empty space. The filesystem backup can be compressed, can
905 contain only part of the encrypted device, but needs to be
906 encrypted separately if so desired.
908 A sector-image will contain the whole partition in encrypted form,
909 for LUKS the LUKS header, the keys-slots and the data area. It can
910 be done under Linux e.g. with dd_rescue (for a direct image copy)
911 and with "cat" or "dd". Example:
913 cat /dev/sda10 > sda10.img
914 dd_rescue /dev/sda10 sda10.img
916 You can also use any other backup software that is capable of making
917 a sector image of a partition. Note that compression is
918 ineffective for encrypted data, hence it does not make sense to
921 For a filesystem backup, you decrypt and mount the encrypted
922 partition and back it up as you would a normal filesystem. In this
923 case the backup is not encrypted, unless your encryption method
924 does that. For example you can encrypt a backup with "tar" as
927 tar cjf - <path> | gpg --cipher-algo AES -c - > backup.tbz2.gpg
929 And verify the backup like this if you are at "path":
931 cat backup.tbz2.gpg | gpg - | tar djf -
933 Note: Allways verify backups, especially encrypted ones.
935 In both cases GnuPG will ask you interactively for your symmetric
936 key. The verify will only output errors. Use "tar dvjf -" to get
937 all comparison results. To make sure no data is written to disk
938 unencrypted, turn off swap if it is not encrypted before doing the
941 You can of course use different or no compression and you can use
942 an asymmetric key if you have one and have a backup of the secret
943 key that belongs to it.
945 A second option for a filestem-level backup that can be used when
946 the backup is also on local disk (e.g. an external USB drive) is
947 to use a LUKS container there and copy the files to be backed up
948 between both mounted containers. Also see next item.
951 * Do I need a backup of the full partition? Would the header and
952 key-slots not be enough?
954 Backup protects you against two things: Disk loss or corruption
955 and user error. By far the most questions on the dm-crypt mailing
956 list about how to recover a damaged LUKS partition are related
957 to user error. For example, if you create a new filesystem on a
958 LUKS partition, chances are good that all data is lost
961 For this case, a header+key-slot backup would often be enough. But
962 keep in mind that a well-treated (!) HDD has roughly a failure
963 risk of 5% per year. It is highly advisable to have a complete
964 backup to protect against this case.
967 * *What do I need to backup if I use "decrypt_derived"?
969 This is a script in Debian, intended for mounting /tmp or swap with
970 a key derived from the master key of an already decrypted device.
971 If you use this for an device with data that should be persistent,
972 you need to make sure you either do not lose access to that master
973 key or have a backup of the data. If you derive from a LUKS
974 device, a header backup of that device would cover backing up the
975 master key. Keep in mind that this does not protect against disk
978 Note: If you recreate the LUKS header of the device you derive from
979 (using luksFormat), the master key changes even if you use the same
980 passphrase(s) and you will not be able to decrypt the derived
981 device with the new LUKS header.
984 * Does a backup compromise security?
986 Depends on how you do it. However if you do not have one, you are
987 going to eventually loseyour encrypted data.
989 There are risks introduced by backups. For example if you
990 change/disable a key-slot in LUKS, a binary backup of the partition
991 will still have the old key-slot. To deal with this, you have to
992 be able to change the key-slot on the backup as well, securely
993 erase the backup or do a filesystem-level backup instead of a binary
996 If you use dm-crypt, backup is simpler: As there is no key
997 management, the main risk is that you cannot wipe the backup when
998 wiping the original. However wiping the original for dm-crypt
999 should consist of forgetting the passphrase and that you can do
1000 without actual access to the backup.
1002 In both cases, there is an additional (usually small) risk with
1003 binary backups: An attacker can see how many sectors and which
1004 ones have been changed since the backup. To prevent this, use a
1005 filesystem level backup methid that encrypts the whole backup in
1006 one go, e.g. as described above with tar and GnuPG.
1008 My personal advice is to use one USB disk (low value data) or
1009 three disks (high value data) in rotating order for backups, and
1010 either use independent LUKS partitions on them, or use encrypted
1011 backup with tar and GnuPG.
1013 If you do network-backup or tape-backup, I strongly recommend to
1014 go the filesystem backup path with independent encryption, as you
1015 typically cannot reliably delete data in these scenarios,
1016 especially in a cloud setting. (Well, you can burn the tape if it
1017 is under your control...)
1020 * What happens if I overwrite the start of a LUKS partition or damage
1021 the LUKS header or key-slots?
1023 There are two critical components for decryption: The salt values
1024 in the header itself and the key-slots. If the salt values are
1025 overwritten or changed, nothing (in the cryptographically strong
1026 sense) can be done to access the data, unless there is a backup
1027 of the LUKS header. If a key-slot is damaged, the data can still
1028 be read with a different key-slot, if there is a remaining
1029 undamaged and used key-slot. Note that in order to make a key-slot
1030 unrecoverable in a cryptographically strong sense, changing about
1031 4-6 bits in random locations of its 128kiB size is quite enough.
1034 * What happens if I (quick) format a LUKS partition?
1036 I have not tried the different ways to do this, but very likely you
1037 will have written a new boot-sector, which in turn overwrites the
1038 LUKS header, including the salts, making your data permanently
1039 irretrivable, unless you have a LUKS header backup. You may also
1040 damage the key-slots in part or in full. See also last item.
1043 * What does the on-disk structure of dm-crypt look like?
1045 There is none. dm-crypt takes a block device and gives encrypted
1046 access to each of its blocks with a key derived from the passphrase
1047 given. If you use a cipher different than the default, you have to
1048 specify that as a parameter to cryptsetup too. If you want to
1049 change the password, you basically have to create a second
1050 encrypted device with the new passphrase and copy your data over.
1051 On the plus side, if you accidentally overwrite any part of a
1052 dm-crypt device, the damage will be limited to the are you
1056 * What does the on-disk structure of LUKS look like?
1058 A LUKS partition consists of a header, followed by 8 key-slot
1059 descriptors, followed by 8 key slots, followed by the encrypted
1062 Header and key-slot descriptors fill the first 592 bytes. The
1063 key-slot size depends on the creation parameters, namely on the
1064 number of anti-forensic stripes, key material offset and master
1067 With the default parameters, each key-slot is a bit less than
1068 128kiB in size. Due to sector alignment of the key-slot start,
1069 that means the key block 0 is at offset 0x1000-0x20400, key
1070 block 1 at offset 0x21000-0x40400, and key block 7 at offset
1071 0xc1000-0xe0400. The space to the next full sector address is
1072 padded with zeros. Never used key-slots are filled with what the
1073 disk originally contained there, a key-slot removed with
1074 "luksRemoveKey" or "luksKillSlot" gets filled with 0xff. Start of
1075 bulk data is at 0x101000, i.e. at 1'052'672 bytes, i.e. at 1MiB
1076 + 4096 bytes from the start of the partition. This is also the
1077 value given by command "luksDump" with "Payload offset: 2056",
1078 just multiply by the sector size (512 bytes). Incidentally,
1079 "luksHeaderBackup" for a LUKS container created with default
1080 parameters dumps exactly the first 1'052'672 bytes to file and
1081 "luksHeaderRestore" restores them.
1083 For non-default parameters, you have to figure out placement
1084 yourself. "luksDump" helps. For the most common non-default
1085 settings, namely aes-xts-plain with 512 bit key, the offsets are:
1086 1st keyslot 0x1000-0x3f800, 2nd keyslot 0x40000-0x7e000, 3rd
1087 keyslot 0x7e000-0xbd800, ..., and start of bulk data at 0x200000.
1089 The exact specification of the format is here:
1090 http://code.google.com/p/cryptsetup/wiki/Specification
1093 * I think this is overly complicated. Is there an alternative?
1095 Not really. Encryption comes at a price. You can use plain
1096 dm-crypt to simplify things a bit. It does not allow multiple
1097 passphrases, but on the plus side, it has zero on disk description
1098 and if you overwrite some part of a plain dm-crypt partition,
1099 exactly the overwritten parts are lost (rounded up to sector
1103 7. Interoperability with other Disk Encryption Tools
1106 * What is this section about?
1108 Cryptsetup for plain dm-crypt can be used to access a number of
1109 on-disk formats created by tools like loop-aes patched into
1110 losetup. This somtimes works and sometimes does not. This section
1111 collects insights into what works, what does not and where more
1112 information is required.
1114 Additional information may be found in the mailing-list archives,
1115 mentioned at the start of this FAQ document. If you have a
1116 solution working that is not yet documented here and think a wider
1117 audience may be intertested, please email the FAQ maintainer.
1120 * loop-aes: General observations.
1122 One problem is that there are different versions of losetup around.
1123 loop-aes is a patch for losetup. Possible problems and deviations
1124 from cryptsetup option syntax include:
1126 - Offsets specifed in bytes (cryptsetup: 512 byte sectors)
1128 - The need to specify an IV offset
1130 - Encryption mode needs specifying (e.g. "-c twofish-cbc-plain")
1132 - Key size needs specifying (e.g. "-s 128" for 128 bit keys)
1134 - Passphrase hash algorithm needs specifying
1136 Also note that because plain dm-crypt and loop-aes format does not
1137 have metadata, autodetection, while feasible in most cases, would
1138 be a lot of work that nobody really wants to do. If you still have
1139 the old set-up, using a verbosity option (-v) on mapping with the
1140 old tool or having a look into the system logs after setup could
1141 give you the information you need.
1144 * loop-aes patched into losetup on debian 5.x, kernel 2.6.32
1146 In this case, the main problem seems to be that this variant of
1147 losetup takes the offset (-o option) in bytes, while cryptsetup
1148 takes it in sectors of 512 bytes each. Example: The losetupp
1151 losetup -e twofish -o 2560 /dev/loop0 /dev/sdb1
1152 mount /dev/loop0 mountpoint
1156 cryptsetup create -c twofish -o 5 --skip 5 e1 /dev/sdb1
1157 mount /dev/mapper/e1 mountpoint
1160 * loop-aes with 160 bit key
1162 This seems to be sometimes used with twofish and blowfish and
1163 represents a 160 bit ripemed160 hash output padded to 196 bit key
1164 length. It seems the corresponding options for cryptsetup are
1166 --cipher twofish-cbc-null -s 192 -h ripemd160:20
1169 8. Issues with Specific Versions of cryptsetup
1172 * When using the create command for plain dm-crypt with cryptsetup
1173 1.1.x, the mapping is incompatible and my data is not accessible
1176 With cryptsetup 1.1.x, the distro maintainer can define different
1177 default encryption modes for LUKS and plain devices. You can check
1178 these compiled-in defaults using "cryptsetup --help". Moreover, the
1179 plain device default changed because the old IV mode was
1180 vulnerable to a watermarking attack.
1182 If you are using a plain device and you need a compatible mode, just
1183 specify cipher, key size and hash algorithm explicitly. For
1184 compatibility with cryptsetup 1.0.x defaults, simple use the
1187 cryptsetup create -c aes-cbc-plain -s 256 -h ripemd160 <name> <dev>
1189 LUKS stores cipher and mode in the metadata on disk, avoiding this
1193 * cryptsetup on SLED 10 has problems...
1195 SLED 10 is missing an essential kernel patch for dm-crypt, which
1196 is broken in its kernel as a result. There may be a very old
1197 version of cryptsetup (1.0.x) provided by SLED, which should also
1198 not be used anymore as well. My advice would be to drop SLED 10.
1200 A. Contributors In no particular order: