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
41 - Ubuntu as of 4/2011: It seems the installer offers to create
42 LUKS partitions in a way that several people mistook for an offer
43 to activate their existing LUKS partition. The installer gives no
44 or an inadequate warning and will destroy your old LUKS header,
45 causing permanent data loss. See also the section on Backup and
48 This issue has been acknowledged by the Ubuntu dev team, see here:
49 http://launchpad.net/bugs/420080
54 Current FAQ maintainer is Arno Wagner <arno@wagner.name>. Other
55 contributors are listed at the end. If you want to contribute, send
56 your article, including a descriptive headline, to the maintainer,
57 or the dm-crypt mailing list with something like "FAQ ..." in the
58 subject. You can also send more raw information and have me write
59 the section. Please note that by contributing to this FAQ, you
60 accept the license described below.
62 This work is under the "Attribution-Share Alike 3.0 Unported"
63 license, which means distribution is unlimited, you may create
64 derived works, but attributions to original authors and this
65 license statement must be retained and the derived work must be
66 under the same license. See
67 http://creativecommons.org/licenses/by-sa/3.0/ for more details of
70 Side note: I did text license research some time ago and I think
71 this license is best suited for the purpose at hand and creates the
75 * Where is the project website?
77 There is the project website at http://code.google.com/p/cryptsetup/
78 Please do not post questions there, nobody will read them. Use
79 the mailing-list instead.
82 * Is there a mailing-list?
84 Instructions on how to subscribe to the mailing-list are at on the
85 project website. People are generally helpful and friendly on the
88 The question of how to unsubscribe from the list does crop up
89 sometimes. For this you need your list management URL, which is
90 sent to you initially and once at the start of each month. Go to
91 the URL mentioned in the email and select "unsubscribe". This page
92 also allows you to request a password reminder.
94 Alternatively, you can send an Email to dm-crypt-request@saout.de
95 with just the word "help" in the subject or message body. Make sure
96 to send it from your list address.
98 The mailing list archive is here:
99 http://dir.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt
105 * What is the difference between "plain" and LUKS format?
107 Plain format is just that: It has no metadata on disk, reads all
108 paramters from the commandline (or the defaults), derives a
109 master-key from the passphrase and then uses that to de-/encrypt
110 the sectors of the device, with a direct 1:1 mapping between
111 encrypted and decrypted sectors.
113 Primary advantage is high resilience to damage, as one damaged
114 encrypted sector results in exactly one damaged decrypted sector.
115 Also, it is not readily apparent that there even is encrypted data
116 on the device, as an overwrite with crypto-grade randomness (e.g.
117 from /dev/urandom) looks exactly the same on disk.
119 Side-note: That has limited value against the authorities. In
120 civilized countries, they cannot force you to give up a crypto-key
121 anyways. In the US, the UK and dictatorships around the world,
122 they can force you to give up the keys (using imprisonment or worse
123 to pressure you), and in the worst case, they only need a
124 nebulous "suspicion" about the presence of encrypted data. My
125 advice is to either be ready to give up the keys or to not have
126 encrypted data when traveling to those countries, especially when
127 crossing the borders.
129 Disadvantages are that you do not have all the nice features that
130 the LUKS metadata offers, like multiple passphrases that can be
131 changed, the cipher being stored in the metadata, anti-forensic
132 properties like key-slot diffusion and salts, etc..
134 LUKS format uses a metadata header and 8 key-slot areas that are
135 being placed ath the begining of the disk, see below under "What
136 does the LUKS on-disk format looks like?". The passphrases are used
137 to decryt a single master key that is stored in the anti-forensic
140 Advantages are a higher usability, automatic configuration of
141 non-default crypto parameters, defenses against low-entropy
142 passphrases like salting and iterated PBKDF2 passphrase hashing,
143 the ability to change passhrases, and others.
145 Disadvantages are that it is readily obvious there is encrypted
146 data on disk (but see side note above) and that damage to the
147 header or key-slots usually results in permanent data-loss. See
148 below under "6. Backup and Data Recovery" on how to reduce that
149 risk. Also the sector numbers get shifted by the length of the
150 header and key-slots and there is a loss of that size in capacity
151 (1MB+4096B for defaults and 2MB for the most commonly used
152 non-default XTS mode).
155 * Can I encrypt an already existing, non-empty partition to use LUKS?
157 There is no converter, and it is not really needed. The way to do
158 this is to make a backup of the device in question, securely wipe
159 the device (as LUKS device initialization does not clear away old
160 data), do a luksFormat, optionally overwrite the encrypted device,
161 create a new filesystem and restore your backup on the now
162 encrypted device. Also refer to sections "Security Aspects" and
163 "Backup and Data Recovery".
165 For backup, plain GNU tar works well and backs up anything likely
166 to be in a filesystem.
169 * How do I use LUKS with a loop-device?
171 This can be very handy for experiments. Setup is just the same as
172 with any block device. If you want, for example, to use a 100MiB
173 file as LUKS container, do something like this:
175 head -c 100M /dev/zero > luksfile # create empty file
176 losetup /dev/loop0 luksfile # map luksfile to /dev/loop0
177 cryptsetup luksFormat /dev/loop0 # create LUKS on loop device
179 Afterwards just use /dev/loop0 as a you would use a LUKS partition.
180 To unmap the file when done, use "losetup -d /dev/loop0".
183 * When I add a new key-slot to LUKS, it asks for a passphrase but
184 then complains about there not being a key-slot with that
187 That is as intended. You are asked a passphrase of an existing
188 key-slot first, before you can enter the passphrase for the new
189 key-slot. Otherwise you could break the encryption by just adding a
190 new key-slot. This way, you have to know the passphrase of one of
191 the already configured key-slots in order to be able to configure a
195 * How do I read a dm-crypt key from file?
197 Note that the file will still be hashed first, just like keyboard
198 input. Use the --key-file option, like this:
200 cryptsetup create --key-file keyfile e1 /dev/loop0
203 * How do I read a LUKS slot key from file?
205 What you really do here is to read a passphrase from file, just as
206 you would with manual entry of a passphrase for a key-slot. You can
207 add a new passphrase to a free key-slot, set the passphrase of an
208 specific key-slot or put an already configured passphrase into a
209 file. In the last case make sure no trailing newline (0x0a) is
210 contained in the key file, or the passphrase will not work because
211 the whole file is used as input.
213 To add a new passphrase to a free key slot from file, use something
216 cryptsetup luksAddKey /dev/loop0 keyfile
218 To add a new passphrase to a specific key-slot, use something like
221 cryptsetup luksAddKey --key-slot 7 /dev/loop0 keyfile
223 To supply a key from file to any LUKS command, use the --key-file
224 option, e.g. like this:
226 cryptsetup luksOpen --key-file keyfile /dev/loop0 e1
229 * How do I read the LUKS master key from file?
231 The question you should ask yourself first is why you would want to
232 do this. The only legitimate reason I can think of is if you want
233 to have two LUKS devices with the same master key. Even then, I
234 think it would be preferable to just use key-slots with the same
235 passphrase, or to use plain dm-crypt instead. If you really have a
236 good reason, please tell me. If I am convinced, I will add how to
240 * What are the security requirements for a key read from file?
242 A file-stored key or passphrase has the same security requirements
243 as one entered interactively, however you can use random bytes and
244 thereby use bytes you cannot type on the keyboard. You can use any
245 file you like as key file, for example a plain text file with a
246 human readable passphrase. To generate a file with random bytes,
247 use something like this:
249 head -c 256 /dev/random > keyfile
252 * If I map a journaled file system using dm-crypt/LUKS, does it still
253 provide its usual transactional guarantees?
255 As far as I know it does (but I may be wrong), but please note that
256 these "guarantees" are far weaker than they appear to be. For
257 example, you may not get a hard flush to disk surface even on a
258 call to fsync. In addition, the HDD itself may do independent
259 write reordering. Some other things can go wrong as well. The
260 filesystem developers are aware of these problems and typically
261 can make it work anyways. That said, dm-crypt/LUKS should not make
264 Personally, I have several instances of ext3 on dm-crypt and have
265 not noticed any specific problems.
267 Update: I did run into frequent small freezes (1-2 sec) when putting
268 a vmware image on ext3 over dm-crypt. This does indicate that the
269 transactional guarantees are in place, but at a cost. When I went
270 back to ext2, the problem went away. This also seems to have gotten
271 better with kernel 2.6.36 and the reworking of filesystem flush
272 locking. Kernel 2.6.38 is expected to have more improvements here.
275 * Can I use LUKS or cryptsetup with a more secure (external) medium
276 for key storage, e.g. TPM or a smartcard?
278 Yes, see the answers on using a file-supplied key. You do have to
279 write the glue-logic yourself though. Basically you can have
280 cryptsetup read the key from STDIN and write it there with your
281 own tool that in turn gets the key from the more secure key
285 * Can I resize a dm-crypt or LUKS partition?
287 Yes, you can, as neither dm-crypt nor LUKS stores partition size.
288 Whether you should is a different question. Personally I recommend
289 backup, recreation of the encrypted partition with new size,
290 recreation of the filesystem and restore. This gets around the
291 tricky business of resizing the filesystem. Resizing a dm-crypt or
292 LUKS container does not resize the filesystem in it. The backup is
293 really non-optional here, as a lot can go wrong, resulting in
294 partial or complete data loss. Using something like gparted to
295 resize an encrypted partition is slow, but typicaly works. This
296 will not change the size of the filesystem hidden under the
299 You also need to be aware of size-based limitations. The one
300 currently relevant is that aes-xts-plain should not be used for
301 encrypted container sizes larger than 2TiB. Use aes-xts-plain64
308 * My dm-crypt/LUKS mapping does not work! What general steps are
309 there to investigate the problem?
311 If you get a specific error message, investigate what it claims
312 first. If not, you may want to check the following things.
314 - Check that "/dev", including "/dev/mapper/control" is there. If it
315 is missing, you may have a problem with the "/dev" tree itself or
316 you may have broken udev rules.
318 - Check that you have the device mapper and the crypt target in your
319 kernel. The output of "dmsetup targets" should list a "crypt"
320 target. If it is not there or the command fails, add device mapper
321 and crypt-target to the kernel.
323 - Check that the hash-functions and ciphers you want to use are in
324 the kernel. The output of "cat /proc/crypto" needs to list them.
327 * My dm-crypt mapping suddenly stopped when upgrading cryptsetup.
329 The default cipher, hash or mode may have changed (the mode changed
330 from 1.0.x to 1.1.x). See under "Issues With Specific Versions of
334 * When I call cryptsetup from cron/CGI, I get errors about unknown
337 If you get errors about unknown parameters or the like that are not
338 present when cryptsetup is called from the shell, make sure you
339 have no older version of cryptsetup on your system that then gets
340 called by cron/CGI. For example some distributions install
341 cryptsetup into /usr/sbin, while a manual install could go to
342 /usr/local/sbin. As a debugging aid, call "cryptsetup --version"
343 from cron/CGI or the non-shell mechanism to be sure the right
347 * Unlocking a LUKS device takes very long. Why?
349 The iteration time for a key-slot (see Section 5 for an explanation
350 what iteration does) is calculated when setting a passphrase. By
351 default it is 1 second on the machine where the passphrase is set.
352 If you set a passphrase on a fast machine and then unlock it on a
353 slow machine, the unlocking time can be much longer. Also take into
354 account that up to 8 key-slots have to be tried in order to find the
357 If this is problem, you can add another key-slot using the slow
358 machine with the same passphrase and then remove the old key-slot.
359 The new key-slot will have an iteration count adjusted to 1 second
360 on the slow machine. Use luksKeyAdd and then luksKillSlot or
363 However, this operation will not change volume key iteration count
364 (MK iterations in output of "cryptsetup luksDump"). In order to
365 change that, you will have to backup the data in the LUKS
366 container, luksFormat on the slow machine and restore the data.
367 Note that in the original LUKS specification this value was fixed
368 to 10, but it is now derived from the PBKDF2 benchmark as well and
369 set to iterations in 0.125 sec or 1000, whichever is larger.
372 * "blkid" sees a LUKS UUID and an ext2/swap UUID on the same device.
375 Some old versions of cryptsetup have a bug where the header does
376 not get completely wiped during LUKS format and an older ext2/swap
377 signature remains on the device. This confuses blkid.
379 Fix: Wipe the unused header areas by doing a backup and restore of
380 the header with cryptsetup 1.1.x:
382 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
383 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
386 * cryptsetup segfaults on Gentoo amd64 hardened ...
388 There seems to be some inteference between the hardening and and
389 the way cryptsetup benchmarks PBKDF2. The solution to this is
390 currently not quite clear for an encrypted root filesystem. For
391 other uses, you can apparently specify USE="dynamic" as compile
392 flag, see http://bugs.gentoo.org/show_bug.cgi?id=283470
398 * Can a bad RAM module cause problems?
400 LUKS and dm-crypt can give the RAM quite a workout, especially when
401 combined with software RAID. In particular the combination RAID5 +
402 LUKS + XFS seems to uncover RAM problems that never caused obvious
403 problems before. Symptoms vary, but often the problem manifest
404 itself when copying large amounts of data, typically several times
405 larger than your main memory.
407 Side note: One thing you should always do on large data
408 copy/movements is to run a verify, for example with the "-d"
409 option of "tar" or by doing a set of MD5 checksums on the source
412 find . -type f -exec md5sum \{\} \; > checksum-file
414 and then a "md5sum -c checksum-file" on the other side. If you get
415 mismatches here, RAM is the primary suspect. A lesser suspect is
416 an overclocked CPU. I have found countless hardware problems in
417 verify runs after copying or making backups. Bit errors are much
418 more common than most people think.
420 Some RAM issues are even worse and corrupt structures in one of the
421 layers. This typically results in lockups, CPU state dumps in the
422 system logs, kernel panic or other things. It is quite possible to
423 have the problem with an encrypted device, but not with an
424 otherwise the same unencrypted device. The reason for that is that
425 encryption has an error amplification property: You flip one bit
426 in an encrypted data block, and the decrypted version has half of
427 its bits flipped. This is an important security property for modern
428 ciphers. With the usual modes in cryptsetup (CBC, ESSIV, XTS), you
429 get up to a completely changed 512 byte block per bit error. A
430 corrupt block causes a lot more havoc than the occasionally
431 flipped single bit and can result in various obscure errors.
433 Note, that a verify run on copying between encrypted or
434 unencrypted devices will reliably detect corruption, even when the
435 copying itself did not report any problems. If you find defect
436 RAM, assume all backups and copied data to be suspect, unless you
442 First you should know that overclocking often makes memory
443 problems worse. So if you overclock (which I strongly recommend
444 against in a system holding data that has some worth), run the
445 tests with the overclocking active.
447 There are two good options. One is Memtest86+ and the other is
448 "memtester" by Charles Cazabon. Memtest86+ requires a reboot and
449 then takes over the machine, while memtester runs from a
450 root-shell. Both use different testing methods and I have found
451 problems fast with each one that the other needed long to find. I
452 recommend running the following procedure until the first error is
455 - Run Memtest86+ for one cycle
457 - Run memterster for one cycle (shut down as many other applications
460 - Run Memtest86+ for 24h or more
462 - Run memtester for 24h or more
464 If all that does not produce error messages, your RAM may be sound,
465 but I have had one weak bit that Memtest86+ needed around 60 hours
466 to find. If you can reproduce the original problem reliably, a good
467 additional test may be to remove half of the RAM (if you have more
468 than one module) and try whether the problem is still there and if
469 so, try with the other half. If you just have one module, get a
470 different one and try with that. If you do overclocking, reduce
471 the settings to the most conservative ones available and try with
478 * Is LUKS insecure? Everybody can see I have encrypted data!
480 In practice it does not really matter. In most civilized countries
481 you can just refuse to hand over the keys, no harm done. In some
482 countries they can force you to hand over the keys, if they suspect
483 encryption. However the suspicion is enough, they do not have to
484 prove anything. This is for practical reasons, as even the presence
485 of a header (like the LUKS header) is not enough to prove that you
486 have any keys. It might have been an experiment, for example. Or it
487 was used as encrypted swap with a key from /dev/random. So they
488 make you prove you do not have encrypted data. Of course that is
489 just as impossible as the other way round.
491 This means that if you have a large set of random-looking data,
492 they can already lock you up. Hidden containers (encryption hidden
493 within encryption), as possible with Truecrypt, do not help
494 either. They will just assume the hidden container is there and
495 unless you hand over the key, you will stay locked up. Don't have
496 a hidden container? Though luck. Anybody could claim that.
498 Still, if you are concerned about the LUKS header, use plain
499 dm-crypt with a good passphrase. See also Section 2, "What is the
500 difference between "plain" and LUKS format?"
503 * Should I initialize (overwrite) a new LUKS/dm-crypt partition?
505 If you just create a filesystem on it, most of the old data will
506 still be there. If the old data is sensitive, you should overwrite
507 it before encrypting. In any case, not initializing will leave the
508 old data there until the specific sector gets written. That may
509 enable an attacker to determine how much and where on the
510 partition data was written. If you think this is a risk, you can
511 prevent this by overwriting the encrypted device (here assumed to
512 be named "e1") with zeros like this:
514 dd_rescue -w /dev/zero /dev/mapper/e1
516 or alternatively with one of the following more standard commands:
518 cat /dev/zero > /dev/mapper/e1
519 dd if=/dev/zero of=/dev/mapper/e1
522 * How do I securely erase a LUKS (or other) partition?
524 For LUKS, if you are in a desperate hurry, overwrite the LUKS
525 header and key-slot area. This means overwriting the first
526 (keyslots x stripes x keysize) + offset bytes. For the default
527 parameters, this is the 1'052'672 bytes, i.e. 1MiB + 4096 of the
528 LUKS partition. For 512 bit key length (e.g. for aes-xts-plain with
529 512 bit key) this is 2MiB. (The diferent offset stems from
530 differences in the sector alignment of the key-slots.) If in doubt,
531 just be generous and overwrite the first 10MB or so, it will likely
532 still be fast enough. A single overwrite with zeros should be
533 enough. If you anticipate being in a desperate hurry, prepare the
534 command beforehand. Example with /dev/sde1 as the LUKS partition
535 and default parameters:
537 head -c 1052672 /dev/zero > /dev/sde1; sync
539 A LUKS header backup or full backup will still grant access to
540 most or all data, so make sure that an attacker does not have
541 access to backups or destroy them as well.
543 If you have time, overwrite the whole LUKS partition with a single
544 pass of zeros. This is enough for current HDDs. For SSDs or FLASH
545 (USB sticks) you may want to overwrite the whole drive several
546 times to be sure data is not retained by wear leveling. This is
547 possibly still insecure as SSD technology is not fully understood
548 in this regard. Still, due to the anti-forensic properties of the
549 LUKS key-slots, a single overwrite of an SSD or FLASH drive could
550 be enough. If in doubt, use physical destruction in addition. Here
551 is a link to some current reseach results on erasing SSDs and FLASH
553 http://www.usenix.org/events/fast11/tech/full_papers/Wei.pdf
555 Keep in mind to also erase all backups.
557 Example for a zero-overwrite erase of partition sde1 done with
560 dd_rescue -w /dev/zero /dev/sde1
563 * How do I securely erase a backup of a LUKS partition or header?
565 That depends on the medium it is stored on. For HDD and SSD, use
566 overwrite with zeros. For an SSD or FLASH drive (USB stick), you
567 may want to overwrite the complete SSD several times and use
568 physical destruction in addition, see last item. For re-writable
569 CD/DVD, a single overwrite should also be enough, due to the
570 anti-forensic properties of the LUKS keyslots. For write-once
571 media, use physical destruction. For low security requirements,
572 just cut the CD/DVD into several parts. For high security needs,
573 shred or burn the medium. If your backup is on magnetic tape, I
574 advise physical destruction by shredding or burning, after
575 overwriting . The problem with magnetic tape is that it has a
576 higher dynamic range than HDDs and older data may well be
577 recoverable after overwrites. Also write-head alignment issues can
578 lead to data not actually being deleted at all during overwrites.
581 * What about backup? Does it compromise security?
583 That depends. See next section.
586 * Why is all my data permanently gone if I overwrite the LUKS header?
588 Overwriting the LUKS header in part or in full is the most common
589 reason why access to LUKS containers is lost permanently.
590 Overwriting can be done in a number of fashions, like creating a
591 new filesystem on the raw LUKS partition, making the raw partition
592 part of a raid array and just writing to the raw partition.
594 The LUKS header contains a 256 bit "salt" value and without that no
595 decryption is possible. While the salt is not secret, it is
596 key-grade material and cannot be reconstructed. This is a
597 cryptographically strong "cannot". From observations on the
598 cryptsetup mailing-list, people typically go though the usual
599 stages of grief (Denial, Anger, Bargaining, Depression, Acceptance)
600 when this happens to them. Observed times vary between 1 day and 2
601 weeks to complete the cycle. Seeking help on the mailing-list is
602 fine. Even if we usually cannot help with getting back your data,
603 most people found the feedback comforting.
605 If your header does not contain an intact salt, best go directly
606 to the last stage ("Acceptance") and think about what to do now.
607 There is one exception that I know of: If your LUKS container is
608 still open, then it may be possible to extract the master key from
609 the running system. Ask on the mailing-list on how to do that and
610 make sure nobody switches off the machine.
615 A salt is a random key-grade value added to the passphrase before
616 it is processed. It is not kept secret. The reason for using salts
617 is as follows: If an attacker wants to crack the password for a
618 single LUKS container, then every possible passphrase has to be
619 tried. Typically an attacker will not try every binary value, but
620 will try words and sentences from a dictionary.
622 If an attacker wants to attack several LUKS containers with the
623 same dictionary, then a different approach makes sense: Compute the
624 resulting slot-key for each dictionary element and store it on
625 disk. Then the test for each entry is just the slow unlocking with
626 the slot key (say 0.00001 sec) instead of calculating the slot-key
627 first (1 sec). For a single attack, this does not help. But if you
628 have more than one container to attack, this helps tremendously,
629 also because you can prepare your table before you even have the
630 container to attack! The calculation is also very simple to
631 parallelize. You could, for example, use the night-time unused CPU
632 power of your desktop PCs for this.
634 This is where the salt comes in. If the salt is combined with the
635 passphrase (in the simplest form, just appended to it), you
636 suddenly need a separate table for each salt value. With a
637 reasonably-sized salt value (256 bit, e.g.) this is quite
641 * Is LUKS secure with a low-entropy (bad) passphrase?
643 This needs a bit of theory. The quality of your passphrase is
644 directly related to its entropy (information theoretic, not
645 thermodynamic). The entropy says how many bits of "uncertainty" or
646 "randomness" are in you passphrase. In other words, that is how
647 difficult guessing the passphrase is.
649 Example: A random English sentence has about 1 bit of entropy per
650 character. A random lowercase (or uppercase) character has about
653 Now, if n is the number of bits of entropy in your passphrase and t
654 is the time it takes to process a passphrase in order to open the
655 LUKS container, then an attacker has to spend at maximum
657 attack_time_max = 2^n * t
659 time for a successful attack and on average half that. There is no
660 way getting around that relationship. However, there is one thing
661 that does help, namely increasing t, the time it takes to use a
662 passphrase, see next FAQ item.
664 Still, if you want good security, a high-entropy passphrase is the
665 only option. Use at least 64 bits for secret stuff. That is 64
666 characters of English text (but only if randomly chosen) or a
667 combination of 12 truly random letters and digits.
669 For passphrase generation, do not use lines from very well-known
670 texts (religious texts, Harry potter, etc.) as they are to easy to
671 guess. For example, the total Harry Potter has about 1'500'000
672 words (my estimation). Trying every 64 character sequence starting
673 and ending at a word boundary would take only something like 20
674 days on a single CPU and is entirely feasible. To put that into
675 perspective, using a number of Amazon EC2 High-CPU Extra Large
676 instances (each gives about 8 real cores), this tests costs
677 currently about 50USD/EUR, but can be made to run arbitrarily fast.
679 On the other hand, choosing 1.5 lines from, say, the Wheel of Time
680 is in itself not more secure, but the book selection adds quite a
681 bit of entropy. (Now that I have mentioned it here, don't use tWoT
682 either!) If you add 2 or 3 typos or switch some words around, then
683 this is good passphrase material.
686 * What is "iteration count" and why is decreasing it a bad idea?
688 Iteration count is the number of PBKDF2 iterations a passphrase is
689 put through before it is used to unlock a key-slot. Iterations are
690 done with the explicit purpose to increase the time that it takes
691 to unlock a key-slot. This provides some protection against use of
692 low-entropy passphrases.
694 The idea is that an attacker has to try all possible passphrases.
695 Even if the attacker knows the passphrase is low-entropy (see last
696 item), it is possible to make each individual try take longer. The
697 way to do this is to repeatedly hash the passphrase for a certain
698 time. The attacker then has to spend the same time (given the same
699 computing power) as the user per try. With LUKS, the default is 1
700 second of PBKDF2 hashing.
702 Example 1: Lets assume we have a really bad passphrase (e.g. a
703 girlfriends name) with 10 bits of entropy. With the same CPU, an
704 attacker would need to spend around 500 seconds on average to
705 break that passphrase. Without iteration, it would be more like
706 0.0001 seconds on a modern CPU.
708 Example 2: The user did a bit better and has 32 chars of English
709 text. That would be about 32 bits of entropy. With 1 second
710 iteration, that means an attacker on the same CPU needs around 136
711 years. That is pretty impressive for such a weak passphrase.
712 Without the iterations, it would be more like 50 days on a modern
713 CPU, and possibly far less.
715 In addition, the attacker can both parallelize and use special
716 hardware like GPUs to speed up the attack. The attack can also
717 happen quite some time after the luksFormat operation and CPUs can
718 have become faster and cheaper. For that reason you want a bit
719 of extra security. Anyways, in Example 1 your are screwed. In
720 example 2, not necessarily. Even if the attack is faster, it still
721 has a certain cost associated with it, say 10000 EUR/USD with
722 iteration and 1 EUR/USD without iteration. The first can be
723 prohibitively expensive, while the second is something you try
724 even without solid proof that the decryption will yield something
727 The numbers above are mostly made up, but show the idea. Of course
728 the best thing is to have a high-entropy passphrase.
730 Would a 100 sec iteration time be even better? Yes and no.
731 Cryptographically it would be a lot better, namely 100 times better.
732 However, usability is a very important factor for security
733 technology and one that gets overlooked surprisingly often. For
734 LUKS, if you have to wait 2 minutes to unlock the LUKS container,
735 most people will not bother and use less secure storage instead. It
736 is better to have less protection against low-entropy passphrases
737 and people actually use LUKS, than having them do without
738 encryption altogether.
740 Now, what about decreasing the iteration time? This is generally a
741 very bad idea, unless you know and can enforce that the users only
742 use high-entropy passphrases. If you decrease the iteration time
743 without ensuring that, then you put your users at increased risk,
744 and considering how rarely LUKS containers are unlocked in a
745 typical work-flow, you do so without a good reason. Don't do it.
746 The iteration time is already low enough that users with entropy
747 low passphrases are vulnerable. Lowering it even further increases
748 this danger significantly.
751 * What about iteration count with plain dm-crypt?
753 Simple: There is none. There is also no salting. If you use plain
754 dm-crypt, the only way to be secure is to use a high entropy
755 passphrase. If in doubt, use LUKS instead.
758 * Is LUKS with default parameters less secure on a slow CPU?
760 Unfortunately, yes. However the only aspect affected is the
761 protection for low-entropy passphrase or master-key. All other
762 security aspects are independent of CPU speed.
764 The master key is less critical, as you really have to work at it
765 to give it low entropy. One possibility is to supply the master key
766 yourself. If that key is low-entropy, then you get what you
767 deserve. The other known possibility is to use /dev/urandom for
768 key generation in an entropy-startved situation (e.g. automatic
769 installation on an embedded device without network and other entropy
772 For the passphrase, don't use a low-entropy passphrase. If your
773 passphrase is good, then a slow CPU will not matter. If you insist
774 on a low-entropy passphrase on a slow CPU, use something like
775 "--iter-time=10" or higher and wait a long time on each LUKS unlock
776 and pray that the attacker does not find out in which way exactly
777 your passphrase is low entropy. This also applies to low-entropy
778 passphrases on fast CPUs. Technology can do only so much to
779 compensate for problems in front of the keyboard.
782 * Why was the default aes-cbc-plain replaced with aes-cbc-essiv?
784 The problem is that cbc-plain has a fingerprint vulnerability, where
785 a specially crafted file placed into the crypto-container can be
786 recognized from the outside. The issue here is that for cbc-plain
787 the initialization vector (IV) is the sector number. The IV gets
788 XORed to the first data chunk of the sector to be encrypted. If you
789 make sure that the first data block to be stored in a sector
790 contains the sector number as well, the first data block to be
791 encrypted is all zeros and always encrypted to the same ciphertext.
792 This also works if the first data chunk just has a constant XOR
793 with the sector number. By having several shifted patterns you can
794 take care of the case of a non-power-of-two start sector number of
797 This mechanism allows you to create a pattern of sectors that have
798 the same first ciphertext block and signal one bit per sector to the
799 outside, allowing you to e.g. mark media files that way for
800 recognition without decryption. For large files this is a
801 practical attack. For small ones, you do not have enough blocks to
802 signal and take care of different file starting offsets.
804 In order to prevent this attack, the default was changed to
805 cbc-essiv. ESSIV uses a keyed hash of the sector number, with the
806 encryption key as key. This makes the IV unpredictable without
807 knowing the encryption key and the watermarking attack fails.
810 * Are there any problems with "plain" IV? What is "plain64"?
812 First, "plain" and "plain64" are both not secure to use with CBC,
813 see previous FAQ item.
815 However there are modes, like XTS, that are secure with "plain" IV.
816 The next limit is that "plain" is 64 bit, with the upper 32 bit set
817 to zero. This means that on volumes larger than 2TiB, the IV
818 repeats, creating a vulnerability that potentially leaks some
819 data. To avoid this, use "plain64", which uses the full sector
820 number up to 64 bit. Note that "plain64" requires a kernel >=
821 2.6.33. Also note that "plain64" is backwards compatible for
822 volume sizes <= 2TiB, but not for those > 2TiB. Finally, "plain64"
823 does not cause any performance penalty compared to "plain".
826 * What about XTS mode?
828 XTS mode is potentially even more secure than cbc-essiv (but only if
829 cbc-essiv is insecure in your scenario). It is a NIST standard and
830 used, e.g. in Truecrypt. At the moment, if you want to use it, you
831 have to specify it manually as "aes-xts-plain", i.e.
833 cryptsetup -c aes-xts-plain luksFormat <device>
835 For volumes >2TiB and kernels >= 2.6.33 use "plain64" (see FAQ
836 item on "plain" and "plain64"):
838 cryptsetup -c aes-xts-plain64 luksFormat <device>
840 There is a potential security issue with XTS mode and large blocks.
841 LUKS and dm-crypt always use 512B blocks and the issue does not
845 6. Backup and Data Recovery
848 * Why do I need Backup?
850 First, disks die. The rate for well-treated (!) disk is about 5%
851 per year, which is high enough to worry about. There is some
852 indication that this may be even worse for some SSDs. This applies
853 both to LUKS and plain dm-crypt partitions.
855 Second, for LUKS, if anything damages the LUKS header or the
856 key-stripe area then decrypting the LUKS device can become
857 impossible. This is a frequent occuurence. For example an
858 accidental format as FAT or some software overwriting the first
859 sector where it suspects a partition boot sector typically makes a
860 LUKS partition permanently inacessible. See more below on LUKS
863 So, data-backup in some form is non-optional. For LUKS, you may
864 also want to store a header backup in some secure location. This
865 only needs an update if you change passphrases.
868 * How do I backup a LUKS header?
870 While you could just copy the appropriate number of bytes from the
871 start of the LUKS partition, the best way is to use command option
872 "luksHeaderBackup" of cryptsetup. This protects also against
873 errors when non-standard parameters have been used in LUKS
874 partition creation. Example:
877 cryptsetup luksHeaderBackup --header-backup-file h /dev/mapper/c1
879 To restore, use the inverse command, i.e.
881 cryptsetup luksHeaderRestore --header-backup-file h /dev/mapper/c1
884 * How do I backup a LUKS or dm-crypt partition?
886 There are two options, a sector-image and a plain file or
887 filesystem backup of the contents of the partition. The sector
888 image is already encrypted, but cannot be compressed and contains
889 all empty space. The filesystem backup can be compressed, can
890 contain only part of the encrypted device, but needs to be
891 encrypted separately if so desired.
893 A sector-image will contain the whole partition in encrypted form,
894 for LUKS the LUKS header, the keys-slots and the data area. It can
895 be done under Linux e.g. with dd_rescue (for a direct image copy)
896 and with "cat" or "dd". Example:
898 cat /dev/sda10 > sda10.img
899 dd_rescue /dev/sda10 sda10.img
901 You can also use any other backup software that is capable of making
902 a sector image of a partition. Note that compression is
903 ineffective for encrypted data, hence it does not make sense to
906 For a filesystem backup, you decrypt and mount the encrypted
907 partition and back it up as you would a normal filesystem. In this
908 case the backup is not encrypted, unless your encryption method
909 does that. For example you can encrypt a backup with "tar" as
912 tar cjf - <path> | gpg --cipher-algo AES -c - > backup.tbz2.gpg
914 And verify the backup like this if you are at "path":
916 cat backup.tbz2.gpg | gpg - | tar djf -
918 Note: Allways verify backups, especially encrypted ones.
920 In both cases GnuPG will ask you interactively for your symmetric
921 key. The verify will only output errors. Use "tar dvjf -" to get
922 all comparison results. To make sure no data is written to disk
923 unencrypted, turn off swap if it is not encrypted before doing the
926 You can of course use different or no compression and you can use a
927 symmetric key if you have one and have a backup of the secret key
930 A second option for a filestem-level backup that can be used when
931 the backup is also on local disk (e.g. an external USB drive) is
932 to use a LUKS container there and copy the files to be backed up
933 between both mounted containers. Also see next item.
936 * Do I need a backup of the full partition? Would the header and
937 key-slots not be enough?
939 Backup protects you against two things: Disk loss or corruption
940 and user error. By far the most questions on the dm-crypt mailing
941 list about how to recover a damaged LUKS partition are related
942 to user error. For example, if you create a new filesystem on a
943 LUKS partition, chances are good that all data is lost
946 For this case, a header+key-slot backup would often be enough. But
947 keep in mind that a well-treated (!) HDD has roughly a failure
948 risk of 5% per year. It is highly advisable to have a complete
949 backup to protect against this case.
952 * *What do I need to backup if I use "decrypt_derived"?
954 This is a script in Debian, intended for mounting /tmp or swap with
955 a key derived from the master key of an already decrypted device.
956 If you use this for an device with data that should be persistent,
957 you need to make sure you either do not lose access to that master
958 key or have a backup of the data. If you derive from a LUKS
959 device, a header backup of that device would cover backing up the
960 master key. Keep in mind that this does not protect against disk
963 Note: If you recreate the LUKS header of the device you derive from
964 (using luksFormat), the master key changes even if you use the same
965 passphrase(s) and you will not be able to decrypt the derived
966 device with the new LUKS header.
969 * Does a backup compromise security?
971 Depends on how you do it. However if you do not have one, you are
972 going to eventually loseyour encrypted data.
974 There are risks introduced by backups. For example if you
975 change/disable a key-slot in LUKS, a binary backup of the partition
976 will still have the old key-slot. To deal with this, you have to
977 be able to change the key-slot on the backup as well, securely
978 erase the backup or do a filesystem-level backup instead of a binary
981 If you use dm-crypt, backup is simpler: As there is no key
982 management, the main risk is that you cannot wipe the backup when
983 wiping the original. However wiping the original for dm-crypt
984 should consist of forgetting the passphrase and that you can do
985 without actual access to the backup.
987 In both cases, there is an additional (usually small) risk with
988 binary backups: An attacker can see how many sectors and which
989 ones have been changed since the backup. To prevent this, use a
990 filesystem level backup methid that encrypts the whole backup in
991 one go, e.g. as described above with tar and GnuPG.
993 My personal advice is to use one USB disk (low value data) or
994 three disks (high value data) in rotating order for backups, and
995 either use independent LUKS partitions on them, or use encrypted
996 backup with tar and GnuPG.
998 If you do network-backup or tape-backup, I strongly recommend to
999 go the filesystem backup path with independent encryption, as you
1000 typically cannot reliably delete data in these scenarios,
1001 especially in a cloud setting. (Well, you can burn the tape if it
1002 is under your control...)
1005 * What happens if I overwrite the start of a LUKS partition or damage
1006 the LUKS header or key-slots?
1008 There are two critical components for decryption: The salt values
1009 in the header itself and the key-slots. If the salt values are
1010 overwritten or changed, nothing (in the cryptographically strong
1011 sense) can be done to access the data, unless there is a backup
1012 of the LUKS header. If a key-slot is damaged, the data can still
1013 be read with a different key-slot, if there is a remaining
1014 undamaged and used key-slot. Note that in order to make a key-slot
1015 unrecoverable in a cryptographically strong sense, changing about
1016 4-6 bits in random locations of its 128kiB size is quite enough.
1019 * What happens if I (quick) format a LUKS partition?
1021 I have not tried the different ways to do this, but very likely you
1022 will have written a new boot-sector, which in turn overwrites the
1023 LUKS header, including the salts, making your data permanently
1024 irretrivable, unless you have a LUKS header backup. You may also
1025 damage the key-slots in part or in full. See also last item.
1028 * What does the on-disk structure of dm-crypt look like?
1030 There is none. dm-crypt takes a block device and gives encrypted
1031 access to each of its blocks with a key derived from the passphrase
1032 given. If you use a cipher different than the default, you have to
1033 specify that as a parameter to cryptsetup too. If you want to
1034 change the password, you basically have to create a second
1035 encrypted device with the new passphrase and copy your data over.
1036 On the plus side, if you accidentally overwrite any part of a
1037 dm-crypt device, the damage will be limited to the are you
1041 * What does the on-disk structure of LUKS look like?
1043 A LUKS partition consists of a header, followed by 8 key-slot
1044 descriptors, followed by 8 key slots, followed by the encrypted
1047 Header and key-slot descriptors fill the first 592 bytes. The
1048 key-slot size depends on the creation parameters, namely on the
1049 number of anti-forensic stripes, key material offset and master
1052 With the default parameters, each key-slot is a bit less than
1053 128kiB in size. Due to sector alignment of the key-slot start,
1054 that means the key block 0 is at offset 0x1000-0x20400, key
1055 block 1 at offset 0x21000-0x40400, and key block 7 at offset
1056 0xc1000-0xe0400. The space to the next full sector address is
1057 padded with zeros. Never used key-slots are filled with what the
1058 disk originally contained there, a key-slot removed with
1059 "luksRemoveKey" or "luksKillSlot" gets filled with 0xff. Start of
1060 bulk data is at 0x101000, i.e. at 1'052'672 bytes, i.e. at 1MiB
1061 + 4096 bytes from the start of the partition. This is also the
1062 value given by command "luksDump" with "Payload offset: 2056",
1063 just multiply by the sector size (512 bytes). Incidentally,
1064 "luksHeaderBackup" for a LUKS container created with default
1065 parameters dumps exactly the first 1'052'672 bytes to file and
1066 "luksHeaderRestore" restores them.
1068 For non-default parameters, you have to figure out placement
1069 yourself. "luksDump" helps. For the most common non-default
1070 settings, namely aes-xts-plain with 512 bit key, the offsets are:
1071 1st keyslot 0x1000-0x3f800, 2nd keyslot 0x40000-0x7e000, 3rd
1072 keyslot 0x7e000-0xbd800, ..., and start of bulk data at 0x200000.
1074 The exact specification of the format is here:
1075 http://code.google.com/p/cryptsetup/wiki/Specification
1078 * I think this is overly complicated. Is there an alternative?
1080 Not really. Encryption comes at a price. You can use plain
1081 dm-crypt to simplify things a bit. It does not allow multiple
1082 passphrases, but on the plus side, it has zero on disk description
1083 and if you overwrite some part of a plain dm-crypt partition,
1084 exactly the overwritten parts are lost (rounded up to sector
1088 7. Interoperability with other Disk Encryption Tools
1091 * What is this section about?
1093 Cryptsetup for plain dm-crypt can be used to access a number of
1094 on-disk formats created by tools like loop-aes patched into
1095 losetup. This somtimes works and sometimes does not. This section
1096 collects insights into what works, what does not and where more
1097 information is required.
1099 Additional information may be found in the mailing-list archives,
1100 mentioned at the start of this FAQ document. If you have a
1101 solution working that is not yet documented here and think a wider
1102 audience may be intertested, please email the FAQ maintainer.
1105 * loop-aes: General observations.
1107 One problem is that there are different versions of losetup around.
1108 loop-aes is a patch for losetup. Possible problems and deviations
1109 from cryptsetup option syntax include:
1111 - Offsets specifed in bytes (cryptsetup: 512 byte sectors)
1113 - The need to specify an IV offset
1115 - Encryption mode needs specifying (e.g. "-c twofish-cbc-plain")
1117 - Key size needs specifying (e.g. "-s 128" for 128 bit keys)
1119 - Passphrase hash algorithm needs specifying
1121 Also note that because plain dm-crypt and loop-aes format does not
1122 have metadata, autodetection, while feasible in most cases, would
1123 be a lot of work that nobody really wants to do. If you still have
1124 the old set-up, using a verbosity option (-v) on mapping with the
1125 old tool or having a look into the system logs after setup could
1126 give you the information you need.
1129 * loop-aes patched into losetup on debian 5.x, kernel 2.6.32
1131 In this case, the main problem seems to be that this variant of
1132 losetup takes the offset (-o option) in bytes, while cryptsetup
1133 takes it in sectors of 512 bytes each. Example: The losetupp
1136 losetup -e twofish -o 2560 /dev/loop0 /dev/sdb1
1137 mount /dev/loop0 mountpoint
1141 cryptsetup create -c twofish -o 5 --skip 5 e1 /dev/sdb1
1142 mount /dev/mapper/e1 mountpoint
1145 * loop-aes with 160 bit key
1147 This seems to be sometimes used with twofish and blowfish and
1148 represents a 160 bit ripemed160 hash output padded to 196 bit key
1149 length. It seems the corresponding options for cryptsetup are
1151 --cipher twofish-cbc-null -s 192 -h ripemd160:20
1154 8. Issues with Specific Versions of cryptsetup
1157 * When using the create command for plain dm-crypt with cryptsetup
1158 1.1.x, the mapping is incompatible and my data is not accessible
1161 With cryptsetup 1.1.x, the distro maintainer can define different
1162 default encryption modes for LUKS and plain devices. You can check
1163 these compiled-in defaults using "cryptsetup --help". Moreover, the
1164 plain device default changed because the old IV mode was
1165 vulnerable to a watermarking attack.
1167 If you are using a plain device and you need a compatible mode, just
1168 specify cipher, key size and hash algorithm explicitly. For
1169 compatibility with cryptsetup 1.0.x defaults, simple use the
1172 cryptsetup create -c aes-cbc-plain -s 256 -h ripemd160 <name> <dev>
1174 LUKS stores cipher and mode in the metadata on disk, avoiding this
1178 * cryptsetup on SLED 10 has problems...
1180 SLED 10 is missing an essential kernel patch for dm-crypt, which
1181 is broken in its kernel as a result. There may be a very old
1182 version of cryptsetup (1.0.x) provided by SLED, which should also
1183 not be used anymore as well. My advice would be to drop SLED 10.
1185 A. Contributors In no particular order: