2 1. General Questions2. Setup3. Common Problems4. Troubleshooting5.
3 Security Aspects6. Backup and Data Recovery7. Issues with Specific
4 Versions of cryptsetupA. Contributors
10 This is the FAQ (Frequently Asked Questions) for cryptsetup. It
11 covers Linux disk encryption with plain dm-crypt (one passphrase,
12 no management, no descriptor on disk) and LUKS (multiple user keys
13 with one master key, anti-forensics, descriptor block at start of
14 device, ...). The latest version should usually be available at
15 http://code.google.com/p/cryptsetup/wiki/FrequentlyAskedQuestions
17 ATTENTION: If you are going to read just one thing, make it the
18 section on Backup and Data Recovery. By far the most questions on
19 the cryptsetup mailing list are from people that just managed to
20 somehow format or overwrite the start of their LUKS partitions. In
21 most cases, there is nothing that can be done to help these poor
22 souls recover their data. Make sure you understand the problem and
23 limitations imposed by the LUKS security model BEFORE you face such
28 Current FAQ maintainer is Arno Wagner <arno@wagner.name>. Other
29 contributors are listed at the end. If you want to contribute, send
30 your article, including a descriptive headline, to the maintainer,
31 or the dm-crypt mailing list with something like "FAQ ..." in the
32 subject. Please note that by contributing to this FAQ, you accept
33 the license described below.
35 This work is under the "Attribution-Share Alike 3.0 Unported"
36 license, which means distribution is unlimited, you may create
37 derived works, but attributions to original authors and this
38 license statement must be retained and the derived work must be
39 under the same license. See
40 http://creativecommons.org/licenses/by-sa/3.0/ for more details of
43 Side note: I did text license research some time ago and I think
44 this license is best suited for the purpose at hand and creates the
47 * Where is the project website?
49 There is the project website at http://code.google.com/p/cryptsetup/
50 Please do not post questions there, nobody will read them. Use
51 the mailing-list instead.
53 * Is there a mailing-list?
55 Instructions on how to subscribe to the mailing-list are at on the
56 project website. People are generally helpful and friendly on the
59 The question of how to unsubscribe from the list does crop up
60 sometimes. For this you need your list management URL, which is
61 sent to you initially and once at the start of each month. Go to
62 the URL mentioned in the email and select "unsubscribe". This page
63 also allows you to request a password reminder.
65 Alternatively, you can send an Email to dm-crypt-request@saout.de
66 with just the word "help" in the subject or message body. Make sure
67 to send it from your list address.
71 * Can I encrypt an already existing, non-empty partition to use
74 There is no converter, and it is not really needed. The way to do
75 this is to make a backup of the device in question, securely wipe
76 the device (as LUKS device initialization does not clear away old
77 data), do a luksFormat, optionally overwrite the encrypted device,
78 create a new filesystem and restore your backup on the now
79 encrypted device. Also refer to sections "Security Aspects" and
80 "Backup and Data Recovery".
82 For backup, plain GNU tar works well and backs up anything likely
83 to be in a filesystem.
85 * How do I use LUKS with a loop-device?
87 Just the same as with any block device. If you want, for example,
88 to use a 100MiB file as LUKS container, do something like this:
90 head -c 100M /dev/zero > luksfile # create empty file
91 losetup /dev/loop0 luksfile # map luksfile to /dev/loop0
92 cryptsetup luksFormat /dev/loop0 # create LUKS on the loop device
94 Afterwards just use /dev/loop0 as a you would use a LUKS partition.
95 To unmap the file when done, use "losetup -d /dev/loop0".
97 * When I add a new key-slot to LUKS, it asks for a passphrase but
98 then complains about there not being a key-slot with that
101 That is as intended. You are asked a passphrase of an existing
102 key-slot first, before you can enter the passphrase for the new
103 key-slot. Otherwise you could break the encryption by just adding a
104 new key-slot. This way, you have to know the passphrase of one of
105 the already configured key-slots in order to be able to configure a
108 * How do I read a dm-crypt key from file?
110 Note that the file will still be hashed first, just like keyboard
111 input. Use the --key-file option, like this:
113 cryptsetup create --key-file keyfile e1 /dev/loop0
115 * How do I read a LUKS slot key from file?
117 What you really do here is to read a passphrase from file, just as
118 you would with manual entry of a passphrase for a key-slot. You can
119 add a new passphrase to a free key-slot, set the passphrase of an
120 specific key-slot or put an already configured passphrase into a
121 file. In the last case make sure no trailing newline (0x0a) is
122 contained in the key file, or the passphrase will not work because
123 the whole file is used as input.
125 To add a new passphrase to a free key slot from file, use something
128 cryptsetup luksAddKey /dev/loop0 keyfile
130 To add a new passphrase to a specific key-slot, use something like
133 cryptsetup luksAddKey --key-slot 7 /dev/loop0 keyfile
135 To supply a key from file to any LUKS command, use the --key-file
136 option, e.g. like this:
138 cryptsetup luksOpen --key-file keyfile /dev/loop0 e1
140 * How do I read the LUKS master key from file?
142 The question you should ask yourself first, is why you would want
143 to do this. The only legitimate reason I can think of is if you
144 want to have two LUKS devices with the same master key. Even then,
145 I think it would be preferable to just use key-slots with the same
146 passphrase, or to use plain dm-crypt instead. If you really have a
147 good reason, please tell me. If I am convinced, I will add how to
150 * What are the security requirements for a key read from file?
152 A file-stored key or passphrase has the same security requirements
153 as one entered interactively, however you can use random bytes and
154 thereby use bytes you cannot type on the keyboard. You can use any
155 file you like as key file, for example a plain text file with a
156 human readable passphrase. To generate a file with random bytes,
157 use something like this:
159 head -c 256 /dev/random > keyfile
161 * If I map a journaled file system using dm-crypt/LUKS, does it
162 still provide its usual transactional guarantees?
164 As far as I know it does (but I may be wrong), but please note that
165 these "guarantees" are far weaker than they appear to be. For
166 example, you not not get a hard flush to disk surface even on a
167 call to fsync. In addition, the HDD itself may do independent
168 write reordering. Some other things can go wrong as well. The
169 filesystem developers are aware of these problems and typically
170 can make it work anyways. That said, dm-crypt/LUKS should not make
173 Personally, I have several instances of ext3 on dm-crypt and have
174 not noticed any specific problems.
176 Update: I did run into frequent small freezes (1-2 sec) when putting
177 a vmware image on ext3 over dm-crypt. This does indicate that the
178 transactional guarantees are in place, but at a cost. When I went
179 back to ext2, the problem went away. This also seems to have gotten
180 better with kernel 2.6.36 and the reworking of filesystem flush
181 locking. Kernel 2.6.37 is expected to improve this even further.
183 * Can I use LUKS or cryptsetup with a more secure (external) medium
184 for key storage, e.g. TPM or a smartcard?
186 Yes, see the answers on using a file-supplied key. You do have to
187 write the glue-logic yourself though. Basically you can have
188 cryptsetup read the key from STDIN and write it there with your
189 own tool that in turn gets the key from the more secure key
192 * Can I resize a dm-crypt or LUKS partition?
194 Yes, you can, as neither dm-crypt nor LUKS stores partition size.
195 Whether you should is a different question. Personally I recommend
196 backup, recreation of the encrypted partition with new size,
197 recreation of the filesystem and restore. This gets around the
198 tricky business of resizing the filesystem. The backup is really
199 non-optional here, as a lot can go wrong, resulting in partial or
200 complete data loss. Using something like gparted to resize an
201 encrypted partition is slow, but pretty safe and should be fine.
202 This will not change the size of the filesystem hidden under the
205 You also need to be aware of size-based limitations. The one
206 currently relevant is that aes-xts-plain should not be used for
207 encrypted container sizes larger than 2TiB. Use aes-xts-plain64
212 * My dm-crypt/LUKS mapping does not work! What general steps are
213 there to investigate the problem?
215 If you get a specific error message, investigate what it claims
216 first. If not, you may want to check the following things.
218 - Check that "/dev", including "/dev/mapper/control" is there. If it
219 is missing, you may have a problem with the "/dev" tree itself or
220 you may have broken udev rules.
222 - Check that you have the device mapper and the crypt target in your
223 kernel. The output of "dmsetup targets" should list a "crypt"
224 target. If it is not there or the command fails, add device mapper
225 and crypt-target to the kernel.
227 - Check that the hash-functions and ciphers you want to use are in
228 the kernel. The output of "cat /proc/crypto" needs to list them.
230 * My dm-crypt mapping suddenly stopped when upgrading cryptsetup.
232 The default cipher, hash or mode may have changed (the mode changed
233 from 1.0.x to 1.1.x). See under "Issues With Specific Versions of
236 * When I call cryptsetup from cron/CGI, I get errors about unknown
239 If you get errors about unknown parameters or the like that are not
240 present when cryptsetup is called from the shell, make sure you
241 have no older version of cryptsetup on your system that then gets
242 called by cron/CGI.For example some distributions install
243 cryptsetup into /usr/sbin, while a manual install could go to
244 /usr/local/sbin. As a debugging aid, call "cryptsetup --version"
245 from cron/CGI or the non-shell mechanism to be sure you have the
248 * Unlocking a LUKS device takes very long. Why?
250 The iteration time for a key-slot (see Section 5 for an explanation
251 what iteration does) is calculated when setting a passphrase. By
252 default it is 1 second on the machine where the passphrase is set.
253 If you set a passphrase on a fast machine and then unlock it on a
254 slow machine, the unlocking time can be much longer. Also take into
255 account that up to 8 key-slots have to be tried in order to find the
258 If this is problem, you can add another key-slot using the slow
259 machine with the same passphrase and then remove the old key-slot.
260 The new key-slot will have an iteration count adjusted to 1 second
261 on the slow machine. Use luksKeyAdd and then luksKillSlot or
264 However, this operation will not change volume key iteration count
265 (MK iterations in output of "cryptsetup luksDump"). In order to
266 change that, you will have to backup the data in the LUKS
267 container, luksFormat on the slow machine and restore the data.
268 Note that in the original LUKS specification this value was fixed
269 to 10, but it is now derived from the PBKDF2 benchmark as well and
270 set to iterations in 0.125 sec or 1000, whichever is larger.
272 * "blkid" sees a LUKS UUID and an ext2/swap UUID on the same device.
275 Some old versions of cryptsetup have a bug where the header does
276 not get completely wiped during LUKS format and an older ext2/swap
277 signature remains on the device. This confuses blkid.
279 Fix: Wipe the unused header areas by doing a backup and restore of
280 the header with cryptsetup 1.1.x:
282 cryptsetup luksHeaderBackup --header-backup-file <file> <device>
283 cryptsetup luksHeaderRestore --header-backup-file <file> <device>
285 If you cannot use a 1.1.x cryptsetup, you can also do a manual wipe
286 of the area in question with the command below. Be very, VERY,
287 careful and make sure to do a backup of the header before. If you
288 get this wrong, your device may become permanently inaccessible.
290 dd if=/dev/zero of=<device> bs=512 seek=2 count=6
292 * cryptsetup segfaults on Gentoo amd64 hardened ...
294 There seems to be some inteference between the hardening and and
295 the way cryptsetup benchmarks PBKDF2. The solution to this is
296 currently not quite clear for an encrypted root filesystem. For
297 other uses, you can apparently specify USE="dynamic" as compile
298 flag, see http://bugs.gentoo.org/show_bug.cgi?id=283470
302 * Can a bad RAM module cause problems?
304 LUKS and dm-crypt can give the RAM quite a workout, especially when
305 combined with software RAID. In particular the combination RAID5 +
306 LUKS + XFS seems to uncover RAM problems that never caused obvious
307 problems before. Symptoms vary, but often the problem manifest
308 itself when copying large amounts of data, typically several times
309 larger than your main memory.
311 Side note: One thing you should always do on large data movements is
312 to run a verify, for example with the "-d" option of "tar" or by
313 doing a set of MD5 checksums on the source or target with
315 find . -type f -exec md5sum \{\} \; > checksum-file
317 and then a "md5sum -c checksum-file" on the other side. If you get
318 mismatches here, RAM is the primary suspect. A lesser suspect is
319 an overclocked CPU. I have found countless hardware problems in
320 verify runs after copying or making backups. Bit errors are much
321 more common than most people think.
323 Some RAM issues are even worse and corrupt structures in one of the
324 layers. This typically results in lockups, CPU state dumps in the
325 system logs, kernel panic or other things. It is quite possible to
326 have the problem with an encrypted device, but not with an
327 otherwise the same unencrypted device. The reason for that is that
328 encryption has an error amplification property: You flip one bit
329 in an encrypted data block, and the decrypted version has half of
330 its bits flipped. This is an important security property for modern
331 ciphers. With the usual modes in cryptsetup (CBC, ESSIV, XTS), you
332 get up to a completely changed 512 byte block per bit error. A
333 corrupt block causes a lot more havoc than the occasionally
334 flipped single bit and can result various obscure errors.
336 Note however that a verify run on copying between encrypted or
337 unencrypted devices can also show you corruption when the copying
338 itself did not report any problems. If you find defect RAM, assume
339 all backups and copied data to be suspect, unless you did a verify.
343 First you should know that overclocking often makes memory problems
344 worse. So if you overclock (which I strongly recommend against in a
345 system holding data that has some worth), run the tests with the
348 There are two good options. One is Memtest86+ and the other is
349 "memtester" by Charles Cazabon. Memtest86+ requires a reboot and
350 then takes over the machine, while memtester runs from a
351 root-shell. Both use different testing methods and I have found
352 problems fast with each one that the other needed long to find. I
353 recommend running the following procedure until the first error is
356 - Run Memtest86+ for one cycle
358 - Run memterster for one cycle (shut down as many other applications
361 - Run Memtest86+ for 24h or more
363 - Run memtester for 24h or more
365 If all that does not produce error messages, your RAM may be sound,
366 but I have had one weak bit that Memtest86+ needed around 60 hours
367 to find. If you can reproduce the original problem reliably, a good
368 additional test may be to remove half of the RAM (if you have more
369 than one module) and try whether the problem is still there and if
370 so, try with the other half. If you just have one module, get a
371 different one and try with that. If you do overclocking, reduce
372 the settings to the most conservative ones available and try with
377 * Should I initialize (overwrite) a new LUKS/dm-crypt partition?
379 If you just create a filesystem on it, most of the old data will
380 still be there. If the old data is sensitive, you should overwrite
381 it before encrypting. In any case, not initializing will leave the
382 old data there until the specific sector gets written. That may
383 enable an attacker to determine how much and where on the
384 partition data was written. If you think this is a risk, you can
385 prevent this by overwriting the encrypted device (here assumed to
386 be named "e1") with zeros like this:
388 dd_rescue -w /dev/zero /dev/mapper/e1
390 or alternatively with one of the following more standard commands:
392 cat /dev/zero > /dev/mapper/e1
393 dd if=/dev/zero of=/dev/mapper/e1
395 * How do I securely erase a LUKS (or other) partition?
397 For LUKS, if you are in a desperate hurry, overwrite the first few
398 kilobytes of the LUKS partition. This erases the master key salt
399 and makes access impossible. However a LUKS header backup or full
400 backup will still grant access to most or all data, so make sure
401 that an attacker does not have access to backups or destroy them as
404 To do this right, overwrite the whole LUKS partition with a single
405 pass of zeros. This is enough for current HDDs. For SSDs you may
406 want to erase the whole drive several times to be sure data is not
407 retained by wear leveling. This is possibly still insecure as SSD
408 technology is not fully understood in this regard. Still, due to
409 the anti-forensic properties of the LUKS key-slots, a single
410 overwrite of an SSD could be enough. If in doubt, use physical
411 destruction in addition. Keep in mind to also erase all backups.
413 Example for a zero-overwrite erase of partition sda10 done with
416 dd_rescue -w /dev/zero /dev/sda10
418 * How do I securely erase a backup of a LUKS partition or header?
420 That depends on the medium it is stored on. For HDD and SSD, use
421 overwrite with zeros. For an SSD, you may want to overwrite the
422 complete SSD several times and use physical destruction in addition,
423 see last item. Treat USB flash drives the same as SSDs. For
424 re-writable CD/DVD, a single overwrite should also be enough, due
425 to the anti-forensic properties of the LUKS keyslots. For
426 write-once media, use physical destruction. For low security
427 requirements, just cut the CD/DVD into several parts. For high
428 security needs, shred or burn the medium. If your backup is on
429 magnetic tape, I advise physical destruction by shredding or
430 burning. The problem with magnetic tape is that it has a higher
431 dynamic range than HDDs and older data may well be recoverable
432 after overwrites. Also write-head alignment issues can lead to
433 data not actually being deleted at all during overwrites.
435 * What about backup? Does it compromise security?
437 That depends. See next section.
439 * Why is all my data gone if I overwrite the LUKS header?
441 Overwriting the LUKS header in part or in full is the most common
442 reason why access to LUKS containers is lost permanently.
443 Overwriting can be done in a number of fashions, like creating a
444 new filesystem on the raw LUKS partition, making the raw partition
445 part of a raid array and just writing to the raw partition.
447 The LUKS header contains a 256 bit "salt" value and without that no
448 decryption is possible. While the salt is not secret, it is
449 key-grade material and cannot be reconstructed. This is a
450 cryptographically strong "cannot". From observations on the
451 cryptsetup mailing-list, people typically go though the usual
452 stages of grief (Denial, Anger, Bargaining, Depression, Acceptance)
453 when this happens to them. Observed times vary between 1 day and 2
454 weeks to complete the cycle. Seeking help on the mailing-list is
455 fine. Even if we usually cannot help with getting back your data,
456 most people found the feedback comforting.
458 If your header does not contain an intact salt, best go directly
459 to the last one ("Acceptance") and think about what to do now.
460 There is one exception that I know of: If your LUKS container is
461 still open, then it may be possible to extract the master key from
462 the running system. Ask on the mailing-list on how to do that and
463 make sure nobody switches off the machine.
467 A salt is a random key-grade value added to the passphrase before
468 it is processed. It is not kept secret. The reason for using salts
469 is as follows: If an attacker wants to crack the password for a
470 single LUKS container, then every possible passphrase has to be
471 tried. Typically an attacker will not try every binary value, but
472 will try words and sentences from a dictionary.
474 If an attacker wants to attack several LUKS containers with the
475 same dictionary, then a different approach makes sense: Compute the
476 resulting slot-key for each dictionary element and store it on
477 disk. Then the test for each entry is just the slow unlocking with
478 the slot key (say 0.00001 sec) instead of calculating the slot-key
479 first (1 sec). For a single attack, this does not help. But if you
480 have more than one container to attack, this helps tremendously,
481 also because you can prepare your table before you even have the
482 container to attack! The calculation is also very simple to
483 parallelize. You could, for example, use the night-time unused CPU
484 power of your desktop PCs for this.
486 This is where the salt comes in. If the salt is combined with the
487 passphrase (in the simplest form, just appended to it), you
488 suddenly need a separate table for each salt value. With a
489 reasonably-sized salt value (256 bit, e.g.) this is quite
492 * Is LUKS secure with a low-entropy (bad) passphrase?
494 This needs a bit of theory. The quality of your passphrase is
495 directly related to its entropy (information theoretic, not
496 thermodynamic). The entropy says how many bits of "uncertainty" or
497 "randomness" are in you passphrase. In other words, that is how
498 difficult guessing the passphrase is.
500 Example: A random English sentence has about 1 bit of entropy per
501 character. A random lowercase (or uppercase) character has about
504 Now, if n is the number of bits of entropy in your passphrase and t
505 is the time it takes to process a passphrase in order to open the
506 LUKS container, then an attacker has to spend at maximum
508 attack_time_max = 2^n * t
510 time for a successful attack and on average half that. There is no
511 way getting around that relationship. However, there is one thing
512 that does help, namely increasing t, the time it takes to use a
513 passphrase, see next FAQ item.
515 Still, if you want good security, a high-entropy passphrase is the
516 only option. Use at least 64 bits for secret stuff. That is 64
517 characters of English text (but only if randomly chosen) or a
518 combination of 12 truly random letters and digits.
520 For passphrase generation, do not use lines from very well-known
521 texts (religious texts, Harry potter, etc.) as they are to easy to
522 guess. For example, the total Harry Potter has about 1'500'000
523 words (my estimation). Trying every 64 character sequence starting
524 and ending at a word boundary would take only something like 20
525 days on a single CPU and is entirely feasible.
527 On the other hand, choosing 1.5 lines from, say, the Wheel of Time
528 is in itself not more secure, but the book selection adds quite a
529 bit of entropy. (Now that I have mentioned it here, don't use tWoT
530 either!) If you add 2 or 3 typos or switch some words around, then
531 this is good passphrase material.
533 * What is "iteration count" and why is decreasing it a bad idea?
535 Iteration count is the number of PBKDF2 iterations a passphrase is
536 put through before it is used to unlock a key-slot. Iterations are
537 done with the explicit purpose to increase the time that it takes
538 to unlock a key-slot. This provides some protection against use of
539 low-entropy passphrases.
541 The idea is that an attacker has to try all possible passphrases.
542 Even if the attacker knows the passphrase is low-entropy (see last
543 item), it is possible to make each individual try take longer. The
544 way to do this is to repeatedly hash the passphrase for a certain
545 time. The attacker then has to spend the same time (given the same
546 computing power) as the user per try. With LUKS, the default is 1
547 second of PBKDF2 hashing.
549 Example 1: Lets assume we have a really bad passphrase (e.g. a
550 girlfriends name) with 10 bits of entropy. With the same CPU, an
551 attacker would need to spend around 500 seconds on average to
552 break that passphrase. Without iteration, it would be more like
553 0.0001 seconds on a modern CPU.
555 Example 2: The user did a bit better and has 32 chars of English
556 text. That would give use about 32 bits of entropy. With 1 second
557 iteration, that means an attacker on the same CPU needs around 136
558 years. That is pretty impressive for such a weak passphrase.
559 Without the iterations, it would be more like 50 days on a modern
560 CPU, and possibly far less.
562 In addition, the attacker can both parallelize and use special
563 hardware like GPUs to speed up the attack. The attack can also
564 happen quite some time after the luksFormat operation and CPUs can
565 have become faster and cheaper. For that reason you want a bit of
566 extra security. Anyways, in Example 1 your are screwed. In example
567 2, not necessarily. Even if the attack is faster, it still has a
568 certain cost associated with it, say 10000 EUR/USD with iteration
569 and 1 EUR/USD without iteration. The first can be prohibitively
570 expensive, while the second is something you try even without
571 solid proof that the decryption will yield something useful.
573 The numbers above are mostly made up, but show the idea. Of course
574 the best thing is to have a high-entropy passphrase.
576 Would a 100 sec iteration time be even better? Yes and no.
577 Cryptographically it would be a lot better, namely 100 times better.
578 However, usability is a very important factor for security
579 technology and one that gets overlooked surprisingly often. For
580 LUKS, if you have to wait 2 minutes to unlock the LUKS container,
581 most people will not bother and use less secure storage instead. It
582 is better to have less protection against low-entropy passphrases
583 and people actually use LUKS, than having them do without
584 encryption altogether.
586 Now, what about decreasing the iteration time? This is generally a
587 very bad idea, unless you know and can enforce that the users only
588 use high-entropy passphrases. If you decrease the iteration time
589 without ensuring that, then you put your users at increased risk,
590 and considering how often LUKS containers are unlocked in a
591 typical work-flow, you do so without a good reason. Don't do it.
592 The iteration time is already low enough that some users will
593 still chose passphrases with entropy low enough that they are
594 vulnerable. Lowering it even further increases this danger
597 * Is LUKS with default parameters less secure on a slow CPU?
599 Unfortunately, yes. However the only aspect affected is the
600 protection for low-entropy passphrase or master-key. All other
601 security aspects are independent of CPU speed.
603 The master key is less critical, as you really have to work at it
604 to give it low entropy. One possibility is to supply the master key
605 yourself. If that key is low-entropy, then you get what you
606 deserve. The other known possibility is to use /dev/urandom for
607 key generation in an entropy-startved situation (e.g. automatic
608 installation on an embedded device without network and other entropy
611 For the passphrase, don't use a low-entropy passphrase. If your
612 passphrase is good, then a slow CPU will not matter. If you insist
613 on a low-entropy passphrase on a slow CPU, use something like
614 "--iter-time=10" or higher and wait a long time on each LUKS unlock
615 and pray that the attacker does not find out in which way exactly
616 your passphrase is low entropy. This also applies to low-entropy
617 passphrases on fast CPUs. Technology can do only so much to
618 compensate for problems in front of the keyboard.
620 * Why was the default aes-cbc-plain replaced with aes-cbc-essiv?
622 The problem is that cbc-plain has a fingerprint vulnerability, where
623 a specially crafted file placed into the crypto-container can be
624 recognized from the outside. The issue here is that for cbc-plain
625 the initialization vector (IV) is the sector number. The IV gets
626 XORed to the first data chunk of the sector to be encrypted. If you
627 make sure that the first data block to be stored in a sector
628 contains the sector number as well, the first data block to be
629 encrypted is all zeros and always encrypted to the same ciphertext.
630 This also works if the first data chunk just has a constant XOR
631 with the sector number. By having several shifted patterns you can
632 take care of the case of a non-power-of-two start sector number of
635 This mechanism allows you to create a pattern of sectors that have
636 the same first ciphertext block and signal one bit per sector to the
637 outside, allowing you to e.g. mark media files that way for
638 recognition without decryption. For large files this is a
639 practical attack. For small ones, you do not have enough blocks to
640 signal and take care of different file starting offsets.
642 In order to prevent this attack, the default was changed to
643 cbc-essiv. ESSIV uses a keyed hash of the sector number, with the
644 encryption key as key. This makes the IV unpredictable without
645 knowing the encryption key and the watermarking attack fails.
647 * Are there any problems with "plain" IV? What is "plain64"?
649 First, "plain" and "plain64" are both not safe to use with CBC, see
652 However there are modes, like XTS, that are secure with "plain" IV.
653 The next limit is that "plain" is 64 bit, with the upper 32 bit set
654 to zero. This means that on volumes larger than 2TiB, the IV
655 repeats, creating a vulnerability that potentially leaks some
656 data. To avoid this, use "plain64", which uses the full sector
657 number up to 64 bit. Note that "plain64" requires a kernel >=
658 2.6.33. Also note that "plain64" is backwards compatible for
659 volume sizes <= 2TiB, but not for those > 2TiB. Finally, "plain64"
660 does not cause any performance penalty compared to "plain".
662 * What about XTS mode?
664 XTS mode is potentially even more secure than cbc-essiv (but only if
665 cbc-essiv is insecure in your scenario). It is a NIST standard and
666 used, e.g. in Truecrypt. At the moment, if you want to use it, you
667 have to specify it manually as "aes-xts-plain", i.e.
669 cryptsetup -c aes-xts-plain luksFormat <device>
671 For volumes >2TiB and kernels >= 2.6.33 use "plain64" (see FAQ
672 item on "plain" and "plain64"):
674 cryptsetup -c aes-xts-plain64 luksFormat <device>
676 There is a potential security issue with XTS mode and large blocks.
677 LUKS and dm-crypt always use 512B blocks and the issue does not
680 6. Backup and Data Recovery
682 * Does a backup compromise security?
684 Depends on how you do it. First, a backup is non-optional with
685 encrypted data just the same way it is with non-encrypted data.
686 Disks do break and they do not care whether they make plain or
687 encrypted data inaccessible.
689 However there are risks introduced by backups. For example if you
690 change/disable a key-slot in LUKS, a binary backup of the partition
691 will still have the old key-slot. To deal with this, you have to
692 be able to change the key-slot on the backup as well, or use a
693 different set-up. One option is to have a different passphrase on
694 the backup and to make the backup with both containers open.
695 Another one is to make a backup of the original, opened container to
696 a single file, e.g. with tar, and to encrypt that file with
697 public-key-cryptography, e.g. with GnuPG. You can then keep the
698 secret key in a safe place, because it is only used to decrypt a
699 backup. The key the backup is encrypted with can be stored without
700 special security measures, as long as an attacker cannot replace
703 If you use dm-crypt, backup is simpler: As there is no key
704 management, the main risk is that you cannot wipe the backup when
705 wiping the original. However wiping the original for dm-crypt
706 should consist of forgetting the passphrase and that you can do
707 without actual access to the backup.
709 In both cases, there is an additional (usually small) risk: An
710 attacker can see how many sectors and which ones have been changed
711 since the backup. This is not possible with the public-key method
714 My personal advice is to use one USB disk (low value date) or three
715 disks (high value data) in rotating order for backups, and either
716 use different passphrases or keep them easily accessible in case
717 you need to disable a key-slot. If you do network-backup or
718 tape-backup, I strongly recommend to go the public-key path,
719 especially as you typically cannot reliably delete data in these
720 scenarios. (Well, you can burn the tape if it is under your
723 * What happens if I overwrite the start of a LUKS partition or
724 damage the LUKS header or key-slots?
726 There are two critical components for decryption: The salt values
727 in the header itself and the key-slots. If the salt values are
728 overwritten or changed, nothing (in the cryptographically strong
729 sense) can be done to access the data, unless there is a backup of
730 the LUKS header. If a key-slot is damaged, the data can still be
731 read with a different key-slot, if there is a remaining undamaged
732 and used key-slot. Note that in order to make a key-slot
733 unrecoverable in a cryptographically strong sense, changing about
734 4-6 bits in random locations of its 128kiB size is quite enough.
736 * What happens if I (quick) format a LUKS partition?
738 I have not tried the different ways to do this, but very likely you
739 will have written a new boot-sector, which in turn overwrites the
740 LUKS header, including the salts. You may also damage the key-slots
741 in part or in full. See also last item.
743 * What does the on-disk structure of dm-crypt look like?
745 There is none. dm-crypt takes a block device and gives encrypted
746 access to each of its blocks with a key derived from the passphrase
747 given. If you use a cipher different than the default, you have to
748 specify that as a parameter to cryptsetup too. If you want to
749 change the password, you basically have to create a second
750 encrypted device with the new passphrase and copy your data over.
751 On the plus side, if you accidentally overwrite any part of a
752 dm-crypt device, the damage will be limited to the are you
755 * What does the on-disk structure of LUKS look like?
757 A LUKS partition consists of a header, followed by 8 key-slot
758 descriptors, followed by 8 key slots, followed by the encrypted
761 Header and key-slot descriptors fill the first 592 bytes. The
762 key-slot size depends on the creation parameters, namely on the
763 number of anti-forensic stripes and on key block alignment.
765 With 4000 stripes (the default), each key-slot is a bit less than
766 128kiB in size. Due to sector alignment of the key-slot start,
767 that means the key block 0 is at offset 0x1000-0x20400, key block
768 1 at offset 0x21000-0x40400, and key block 7 at offset
769 0xc1000-0xe0400. The space to the next full sector address is
770 padded with zeros. Never used key-slots are filled with what the
771 disk originally contained there, a key-slot removed with
772 "luksRemoveKey" or "luksKillSlot" gets filled with 0xff. Start of
773 bulk data (with the default 4000 stripes and 8 key-slots) is at
774 0x101000, i.e. at 1'052'672 bytes, i.e. at 1MiB + 4096 bytes from
775 the start of the partition. This is also the value given by command
776 "luksDump" with "Payload offset: 2056", just multiply by the sector
777 size (512 bytes). Incidentally, "luksHeaderBackup" dumps exactly
778 the first 1'052'672 bytes to file and "luksHeaderRestore" restores
781 The exact specification of the format is here:
782 http://code.google.com/p/cryptsetup/wiki/Specification
784 * How do I backup a LUKS header?
786 While you could just copy the appropriate number of bytes from the
787 start of the LUKS partition, the best way is to use command option
788 "luksHeaderBackup" of cryptsetup. This protects also against errors
789 when non-standard parameters have been used in LUKS partition
793 cryptsetup luksHeaderBackup --header-backup-file h_bak /dev/mapper/c1
795 * How do I backup a LUKS partition?
797 You do a sector-image of the whole partition. This will contain the
798 LUKS header, the keys-slots and the data ares. It can be done
799 under Linux e.g. with dd_rescue (for a direct image copy) and with
800 "cat" or "dd". Example:
802 cat /dev/sda10 > sda10.img
803 dd_rescue /dev/sda10 sda10.img
805 You can also use any other backup software that is capable of making
806 a sector image of a partition. Note that compression is
807 ineffective for encrypted data, hence it does not sense to use it.
809 * Do I need a backup of the full partition? Would the header and
810 key-slots not be enough?
812 Backup protects you against two things: Disk loss or corruption and
813 user error. By far the most questions on the dm-crypt mailing list
814 about how to recover a damaged LUKS partition are related to user
815 error. For example, if you create a new filesystem on a LUKS
816 partition, chances are good that all data is lost permanently.
818 For this case, a header+key-slot backup would often be enough. But
819 keep in mind that a HDD has roughly a failure risk of 5% per year.
820 It is highly advisable to have a complete backup to protect against
823 * Are there security risks from a backup of the LUKS header or a
824 whole LUKS partition?
826 Yes. One risk is that if you remove access rights for specific
827 key-slots by deleting their contents, the data can still be
828 accessed with invalidated passphrase and the backup. The other risk
829 is that if you erase a LUKS partition, a backup could still grant
830 access, especially if you only erased the LUKS header and not the
833 * I think this is overly complicated. Is there an alternative?
835 Yes, you can use plain dm-crypt. It does not allow multiple
836 passphrases, but on the plus side, it has zero on disk description
837 and if you overwrite some part of a plain dm-crypt partition,
838 exactly the overwritten parts are lost (rounded up to sector
841 7. Issues with Specific Versions of cryptsetup
843 * When using the create command for plain dm-crypt with cryptsetup
844 1.1.x, the mapping is incompatible and my data is not accessible
847 With cryptsetup 1.1.x, the distro maintainer can define different
848 default encryption modes for LUKS and plain devices. You can check
849 these compiled-in defaults using "cryptsetup --help". Moreover, the
850 plain device default changed because the old IV mode was
851 vulnerable to a watermarking attack.
853 If you are using a plain device and you need a compatible mode, just
854 specify cipher, key size and hash algorithm explicitly. For
855 compatibility with cryptsetup 1.0.x defaults, simple use the
858 cryptsetup create -c aes-cbc-plain -s 256 -h ripemd160 <name> <device>
860 LUKS stores cipher and mode in the metadata on disk, avoiding this
863 * cryptsetup on SLED 10 has problems...
865 SLED 10 is missing an essential kernel patch for dm-crypt, which
866 is broken in its kernel as a result. There may be a very old
867 version of cryptsetup (1.0.x) provided by SLED, which should also
868 not be used anymore as well. My advice would be to drop SLED 10.
870 A. Contributors In no particular order: