NAME
ccrypt - encrypt and decrypt files and streams
SYNOPSIS
ccrypt [mode] [options] [file...]
ccencrypt [options] [file...]
ccdecrypt [options] [file...]
ccat [options] file...
DESCRIPTION
ccrypt is a utility for encrypting and decrypting files and streams. It
was designed to replace the standard unix crypt utility, which is
notorious for using a very weak encryption algorithm. ccrypt is based
on the Rijndael block cipher, which was also chosen by the U.S.
government as the Advanced Encryption Standard (AES, see
http://www.nist.gov/aes/). This cipher is believed to provide very
strong cryptographic security.
Unlike unix crypt, the algorithm provided by ccrypt is not symmetric,
i.e., one must specify whether to encrypt or decrypt. The most common
way to invoke ccrypt is via the commands ccencrypt and ccdecrypt.
Encryption and decryption depends on a keyword (or key phrase) supplied
by the user. By default, the user is prompted to enter a keyword from
the terminal. Keywords can consist of any number of characters, and all
characters are significant (although ccrypt internally hashes the key
to 256 bits). Longer keywords provide better security than short ones,
since they are less likely to be discovered by exhaustive search.
MODES
ccrypt can operate in five different modes. If more than one mode is
specified, the last one specified takes precedence. The aliases
ccencrypt, ccdecrypt, and ccat are provided as a convenience; they are
equivalent to ccrypt -e, ccrypt -d, and ccrypt -c, respectively.
-e, --encrypt Encrypt. This is the default mode. If filename arguments
are given, encrypt the files and append the suffix .cpt
to their names. Otherwise, run as a filter.
-d, --decrypt Decrypt. If filename arguments are given, decrypt the
files and strip the suffix .cpt from the filenames, if
present. Otherwise, run as a filter.
-c, --cat Decrypt one or more files to standard output. If no
filename arguments are given, decrypt as a filter.
Implies -l.
-x, --keychange
Change the key of encrypted data. In this mode, ccrypt
prompts for two passwords: the old one and the new one.
If filename arguments are given, modify the files.
Otherwise, run as a filter.
-u, --unixcrypt
Simulate the old unix crypt command. Note: the cipher
used by unix crypt has been broken and is not secure.
Please use this option only to decrypt existing files.
If filename arguments are given, decrypt the files to
stdout. Otherwise, run as a filter. Note that for the
unix crypt format, there is no easy way to detect
whether a given key matches or not; thus, for safety,
this mode does not overwrite files.
OPTIONS
The following options are supported in addition to the modes described
above:
-b, --brave When reading an encryption key from the terminal, ask
the user only once for the key. By default, ccrypt will
ask the user to enter such keys twice, as a safeguard
against accidentally destroying data due to a mistyped
key. Using the --brave option disables this safety
check. Never use it, unless you know what you are doing.
See also --timid.
-E var, --envvar var
Read the keyword from the environment variable var.
Note that this might be unsafe on certain systems,
namely where users can use the ps command to see the
environment of a process owner by another user. On most
modern systems, however, such behavior of ps is disabled
and the -E option should be safe there. Also, as an
added security measure, ccrypt erases the keyword from
its environment immediately after reading it.
-f, --force Overwrite existing files or operate on write-protected
files without asking any questions. Also, override
ccrypt’s reluctance to write or read encrypted data to
or from a terminal.
-F var, --envvar2 var
Same as -E, except for second keyword (in keychange
mode).
-h, --help Help. Print usage information and exit.
-H key, --key2 key
Same as -K, except for second keyword (in keychange
mode).
-k file, --keyfile file
Read the keyword as the first line from the named file.
In keychange mode, two keywords are read as the first
two lines of the file. The filename "-" may be given for
reading keywords from standard input. Using the -k -
option and sending the keyword on stdin is probably the
safest way to pass a keyword to ccrypt from another
program or script.
-K key, --key key
Specify the keyword on the command line. This is unsafe,
because any other user can see the command line by
running the ps command. Only use this option for
testing purposes, and never with a real keyword.
-l, --symlinks Force encryption/decryption of symbolic links. By
default, symbolic links are ignored except in cat or
unixcrypt mode. Note that with the -l option,
encryption/decryption of a symbolic link causes the
suffix .cpt to be added/removed from the name of the
link, not the name of the file pointed to.
-L, --license Print license info and exit.
-m, --mismatch Normally, ccrypt refuses to decrypt data with a key that
does not seem to match. The -m option overrides this
restriction. This can sometimes be useful in recovering
data from a corrupted file (see RECOVERING DATA FROM
CORRUPTED FILES). To avoid irretrievable loss of data
when decrypting with a wrong key, this option cannot be
used with modes that overwrite the input file.
-P prompt, --prompt prompt
Use prompt instead of the default prompt "Enter
encryption/decryption key: ". This may be useful in
some shell scripts.
-q, --quiet Suppress most warnings.
-Q prompt, --prompt2 prompt
Same as -P, except for second keyword (in keychange
mode).
-r, --recursive
Traverse subdirectories recursively.
-R, --rec-symlinks
Traverse subdirectories recursively, and also follow
symbolic links to subdirectories.
-s, --strictsuffix
Refuse to encrypt files which already have the .cpt
suffix (or that selected with -S). This can be useful
when adding some files to a directory of already
encrypted files. This option has no effect in decryption
or keychange mode.
-S .suf, --suffix .suf
Use the suffix .suf instead of the default suffix .cpt.
-t, --timid When reading an encryption key from the terminal, ask
the user to enter the key twice. If the two entered keys
are not identical, abort. This is a safeguard against
accidentally destroying data by encrypting it with a
mistyped key. Note: this behavior is now the default,
and can be overridden with the --brave option.
-T, --tmpfiles This option causes ccrypt to use temporary files during
encryption/decryption, rather than overwriting the file
contents destructively. This method leaves the original
file contents lying around in unused sectors of the file
system, and thus is less secure than the default
behavior. However, in situations where this loss of
security is not important, the --tmpfiles option can
provide a measure of protection against data being
corrupted due to a system crash in the middle of
overwriting a file.
-v, --verbose Print progress information to stderr.
-V, --version Print version info and exit.
-- End of options. Any remaining arguments are interpreted
as filenames. This also turns off filter mode, even if
zero filenames follow. This might be useful in the
context of shell pattern expansion; ccrypt -- * will
behave correctly even if no files match the pattern *.
NOTES ON USAGE
The user interface of ccrypt intentionally resembles that of GNU gzip,
although it is not identical. When invoked with filename arguments,
ccrypt normally modifies the files in place, overwriting their old
content. Unlike gzip, the output is not first written to a temporary
file; instead, the data is literally overwritten. For encryption, this
is usually the desired behavior, since one does not want copies of the
unencrypted data to remain in hidden places in the file system. The
disadvantage is that if ccrypt is interrupted in the middle of writing
to a file, the file will end up in a corrupted, partially encrypted
state. However, in such cases it is possible to recover most of the
data; see RECOVERING DATA FROM CORRUPTED FILES below. If you want to
force ccrypt to use temporary files, use the --tmpfiles option.
When ccrypt receives an interrupt signal (Ctrl-C) while updating a file
in place, it does not exit immediately, but rather delays the exit
until after it finishes writing to the current file. This is to prevent
files from being partially overwritten and thus corrupted. If you want
to force ccrypt to exit immediately, just press Ctrl-C twice quickly.
The encryption algorithm used by ccrypt uses a random seed which is
different each time. As a result, encrypting the same file twice will
never yield the same result. The advantage of this method is that
similarities in plaintext do not lead to similarities in ciphertext;
there is no way of telling whether the content of two encrypted files
is similar or not.
Because of the use of a random seed, decrypting and re-encrypting a
file with the same key will not lead to an identical file. It is
primarily for this reason that ccrypt refuses to decrypt files with a
non-matching key; if this were allowed, there would be no way
afterwards to restore the original file, and the data would be
irretrievably lost.
When overwriting files, special care is taken with hard links and
symbolic links. Each physical file (i.e., each inode) is processed at
most once, no matter how many paths to it are encountered on the
command line or in subdirectories traversed recursively. For each file
which has multiple hard links, a warning is printed, to alert the user
that not all paths to the file might have been properly renamed.
Symbolic links are ignored except in cat mode, or unless the -l or -R
option is given.
Unlike gzip, ccrypt does not complain about files that have improper
suffixes. It is legal to doubly encrypt a file. It is also legal to
decrypt a file which does not have the .cpt suffix, provided the file
contains valid data for the given decryption key. Use the
--strictsuffix option if you want to prevent ccrypt from encrypting
files which already have a .cpt suffix.
Regarding encryption and compression: encrypted data is statistically
indistinguishable from random data, and thus it cannot be compressed.
But of course it is possible to compress the data first, then encrypt
it. Suggested file suffixes are .gz.cpt or .gzc.
RECOVERING DATA FROM CORRUPTED FILES
Encrypted data might be corrupted for a number of reasons. For
instance, a file might have been partially encrypted or decrypted if
ccrypt was interrupted while processing the file. Or data might be
corrupted by a software or hardware error, or during transmission over
a network. The encryption algorithm used by ccrypt is designed to allow
recovery from errors. In general, only a few bytes of data will be lost
near where the error occurred.
Data encrypted by ccrypt can be thought of as a sequence of 32-byte
blocks. To decrypt a particular block, ccrypt only needs to know the
decryption key, the data of the block itself, and the data of the block
immediately preceding it. ccrypt cannot tell whether a block is
corrupted or not, except the very first block, which is special. Thus,
if the encrypted data has been altered in the middle or near the end of
a file, ccrypt can be run to decrypt it as usual, and most of the data
will be decrypted correctly, except near where the corruption occurred.
The very first block of encrypted data is special, because it does not
actually correspond to any plaintext data; this block holds the random
seed generated at encryption time. ccrypt also uses the very first
block to decide whether the given keyword matches the data or not. If
the first block has been corrupted, ccrypt will likely decide that the
keyword does not match; in such cases, the -m option can be used to
force ccrypt to decrypt the data anyway.
If a file contains some encrypted and some unencrypted data, or data
encrypted with two different keys, one should decrypt the entire file
with each applicable key, and then piece together the meaningful parts
manually.
Finally, decryption will only produce meaningful results if the data is
aligned correctly along block boundaries. If the block boundary
information has been lost, one has to try all 32 possibilities.
DESCRIPTION OF THE CIPHER
Block ciphers operate on data segments of a fixed length. For
instance, the Rijndael block cipher used in ccrypt has a block length
of 32 bytes or 256 bits. Thus, this cipher encrypts 32 bytes at a time.
Stream ciphers operate on data streams of any length. There are several
standard modes for operating a block cipher as a stream cipher. One
such standard is Cipher Feedback (CFB), defined in FIPS 81 and ANSI
X3.106-1983. ccrypt implements a stream cipher by operating the
Rijndael block cipher in CFB mode.
Let P[i] and C[i] be the ith block of the plaintext and ciphertext,
respectively. CFB mode specifies that
C[i] = P[i] ^ E(k,C[i-1])
Here ^ denotes the bitwise exclusive or function, and E(k,x) denotes
the encryption of the block x under the key k using the block cipher.
Thus, each block of the ciphertext is calculated from the corresponding
block of plaintext and the previous block of ciphertext. Note that in
fact, each byte of P[i] can be calculated from the corresponding byte
of C[i], so that the stream cipher can be applied to one byte at a
time. In particular, the stream length need not be a multiple of the
block size.
Assuming that blocks are numbered starting from 0, a special "initial"
ciphertext block C[-1] is needed to provide the base case for the above
formula. This value C[-1] is called the initialization vector or seed.
The seed is chosen at encryption time and written as the first block of
the encrypted stream. It is important never to use the same seed more
than once; otherwise, the two resulting ciphertext blocks C[0] could be
related by a simple xor to obtain information about the corresponding
plaintext blocks P[0]. If the same seed is never reused, CFB is
provably as secure as the underlying block cipher.
In ccrypt, the seed is constructed as follows: first, a combination of
the host name, current time, process id, and an internal counter are
hashed into a 28-byte value, using a cryptographic hash function. A
fixed four-byte "magic number" is combined with this value, and the
resulting 32-byte value is encrypted by one round of the Rijndael block
cipher with the given key. The result is used as the seed and appended
to the beginning of the ciphertext. The use of the magic number allows
ccrypt to detect non-matching keys before decryption.
SECURITY
ccrypt is believed to provide very strong cryptographic security,
equivalent to that of the Rijndael cipher with 256-bit block size and
256-bit key size. Another version of the Rijndael cipher (with a
smaller block size) is used in the U.S. government’s Advanced
Encryption Standard (AES, see http://www.nist.gov/aes/). Therefore,
this cipher is very well studied and subject to intensive public
scrutiny. This scrutiny has a positive effect on the cipher’s
security. In particular, if an exploitable weakness in this cipher were
ever discovered, this would become widely publicized.
In practical terms, the security of ccrypt means that, without
knowledge of the encryption key, it is effectively impossible to obtain
any information about the plaintext from a given ciphertext. This is
true even if a large number of plaintext-ciphertext pairs are already
known for the same key. Moreover, because ccrypt uses a key size of 256
bits, an exhaustive search of the key space is not feasible, at least
as long as sufficiently long keys are actually used in practice. No
cipher is secure if users choose insecure keywords.
On the other hand, ccrypt does not attempt to provide data integrity,
i.e., it will not attempt to detect whether the ciphertext was modified
after encryption. In particular, encrypted data can be truncated,
leaving the corresponding decrypted data also truncated, but otherwise
consistent. If one needs to ensure data integrity as well as secrecy,
this can be achieved by other methods. The recommended method is to
prepend a cryptographic hash (for instance, an MD5 or SHA-1 hash) to
the data before encryption.
ccrypt does not claim to provide any particular safeguards against
information leaking via the local operating system. While reasonable
precautions are taken, there is no guarantee that keywords and
plaintexts have been physically erased after encryption in completed;
parts of such data might still exist in memory or on disk. ccrypt does
not currently use priviledged memory pages.
When encrypting files, ccrypt by default accesses them in read-write
mode. This normally causes the original file to be physically
overwritten, but on some file systems, this might not be the case.
Note that the use of the -K option is unsafe in a multiuser
environment, because the command line of a process is visible to other
users running the ps command. The use of the -E option is potentially
unsafe for the same reason, although recent versions of ps don’t tend
to display environment information to other users. The use of the -T
option is unsafe for encryption because the original plaintext will
remain in unused sectors of the file system.
EMACS PACKAGE
There is an emacs package for reading and writing encrypted files.
(Note that this package currently only works with emacs, not with
xemacs.) This package hooks into the low-level file I/O functions of
emacs, prompting the user for a password where appropriate. It is
implemented in much the same way as support for compressed files; in
fact, the package, whose name is jka-compr-ccrypt, is based directly on
the jka-compr package which is part of GNU Emacs. It handles both
encrypted and compressed files (although currently not encrypted
compressed files).
To use the package, simply load jka-compr-ccrypt, then edit as usual.
When you open a file with the ".cpt" extension, emacs will prompt you
for a password for the file. It will remember the password for the
buffer, and when you save the file later, it will be automatically
encrypted again (provided you save it with a ".cpt" extension). Except
for the password prompt, the operation of the package should be
transparent to the user. The package also handles compressed ".gz",
".bz2", and ".Z" files, and it should be used instead of, not in
addition to, jka-compr. The command M-x ccrypt-set-buffer-password can
be used to change the current password of a buffer.
The simplest way to use this package is to include the lines
(setq load-path (cons "path" load-path))
(require ’jka-compr-ccrypt "jka-compr-ccrypt.el")
in your .emacs file, where path is the directory which holds the file
jka-compr-ccrypt.el. You may also need to disable loading of the old
jka-compr package, since the two packages are not compatible with each
other (and in any case, jka-compr-ccrypt contains all the functionality
of jka-compr).
Limitations of the emacs package: there is no guarantee that
unencrypted information cannot leak to the file system; in fact, the
package sometimes writes unencrypted data to temporary files. However,
auto-saved files are normally treated correctly (i.e., encrypted). For
details, see the comments in the file jka-compr-ccrypt.el.
EXIT STATUS
The exit status is 0 on successful completion, and non-zero otherwise.
An exit status of 1 means illegal command line, 2 is out of memory, 3
is a fatal i/o error, 4 is a non-matching key or wrong file format, 6
is interrupt, 7 is mistyped key in --timid mode, and 8 is a non-fatal
i/o error.
Fatal i/o errors are those which occur while processing a file which is
already open. Such errors cause ccrypt to abort its operation
immediately with an exit status of 3. Non-fatal i/o errors are those
which occur while handling files which are not already open; typically,
such errors are caused by files that are missing, not readable, or
can’t be created. When encountering a non-fatal i/o error, ccrypt
simply continues to process the next available input file. The exit
status of 8 is delayed until after all the files have been processed.
Non-matching keys and wrong file formats are also considered non-fatal
errors, and cause ccrypt to continue with processing the next available
input file. In this case, an exit status of 4 is given after all the
files have been processed. If there is a conflict between exit status 4
and 8, then 8 is returned.
The former exit status 5 ("wrong file format") has been eliminated, and
is now covered under exit status 4 ("non-matching key or wrong file
format"). Note that ccrypt does not really have a "file format" in the
proper sense of the word; any file of length at least 32 bytes is
potentially a valid encrypted file.
BUGS
While ccrypt can handle keywords of arbitrary length, some operating
systems limit the length of an input line to 1024 characters.
The renaming of files (adding or removing the .cpt suffix) can go wrong
if a filename is repeated on the command line. In this case, the file
is only encrypted/decrypted once, but the suffix may be added or
removed several times. This is because ccrypt thinks it encountered
different hardlinks for the same file.
The --strictsuffix option can behave in unexpected ways if one file has
several hardlinks of which some have the suffix and some don’t. In this
case, the inode will be encrypted/decrypted, but the suffix will be
changed only for those filenames that allow it. Similarly, if a file
cannot be renamed because a file of the given name already exists, the
file may still be encrypted/decrypted if it has another hardlink.
VERSION
1.7
AUTHOR
Peter Selinger <selinger at users.sourceforge.net>
COPYRIGHT
Copyright (C) 2000-2004 Peter Selinger
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
59 Temple Place, Suite 330, Boston, MA 02111-1307, USA. See also
http://www.gnu.org/.