Allow the planner's estimate of the fraction of a cursor's rows that will be

[postgresql] / doc / src / sgml / pgcrypto.sgml

<!-- $PostgreSQL: pgsql/doc/src/sgml/pgcrypto.sgml,v 1.6 2008/01/17 14:34:45 mha Exp $ -->

<sect1 id="pgcrypto">
 <title>pgcrypto</title>

 <indexterm zone="pgcrypto">
  <primary>pgcrypto</primary>
 </indexterm>

 <para>
  The <filename>pgcrypto</> module provides cryptographic functions for
  <productname>PostgreSQL</>.
 </para>

 <sect2>
  <title>General hashing functions</title>

  <sect3>
   <title><function>digest()</function></title>

   <synopsis>
    digest(data text, type text) returns bytea
    digest(data bytea, type text) returns bytea
   </synopsis>

   <para>
    Computes a binary hash of the given <parameter>data</>.
    <parameter>type</> is the algorithm to use.
    Standard algorithms are <literal>md5</literal>, <literal>sha1</literal>,
    <literal>sha224</literal>, <literal>sha256</literal>,
    <literal>sha384</literal> and <literal>sha512</literal>.
    If <filename>pgcrypto</> was built with
    OpenSSL, more algorithms are available, as detailed in
    <xref linkend="pgcrypto-with-without-openssl">.
   </para>

   <para>
    If you want the digest as a hexadecimal string, use
    <function>encode()</> on the result.  For example:
   </para>
   <programlisting>
    CREATE OR REPLACE FUNCTION sha1(bytea) returns text AS $$
      SELECT encode(digest($1, 'sha1'), 'hex')
    $$ LANGUAGE SQL STRICT IMMUTABLE;
   </programlisting>
  </sect3>

  <sect3>
   <title><function>hmac()</function></title>

   <synopsis>
    hmac(data text, key text, type text) returns bytea
    hmac(data bytea, key text, type text) returns bytea
   </synopsis>

   <para>
    Calculates hashed MAC for <parameter>data</> with key <parameter>key</>.
    <parameter>type</> is the same as in <function>digest()</>.
   </para>

   <para>
    This is similar to <function>digest()</> but the hash can only be
    recalculated knowing the key.  This prevents the scenario of someone
    altering data and also changing the hash to match.
   </para>

   <para>
    If the key is larger than the hash block size it will first be hashed and
    the result will be used as key.
   </para>
  </sect3>
 </sect2>

 <sect2>
  <title>Password hashing functions</title>

  <para>
   The functions <function>crypt()</> and <function>gen_salt()</>
   are specifically designed for hashing passwords.
   <function>crypt()</> does the hashing and <function>gen_salt()</>
   prepares algorithm parameters for it.
  </para>

  <para>
   The algorithms in <function>crypt()</> differ from usual hashing algorithms
   like MD5 or SHA1 in the following respects:
  </para>

  <orderedlist>
   <listitem>
    <para>
     They are slow.  As the amount of data is so small, this is the only
     way to make brute-forcing passwords hard.
    </para>
   </listitem>
   <listitem>
    <para>
     They use a random value, called the <firstterm>salt</>, so that users
     having the same password will have different encrypted passwords.
     This is also an additional defense against reversing the algorithm.
    </para>
   </listitem>
   <listitem>
    <para>
     They include the algorithm type in the result, so passwords hashed with
     different algorithms can co-exist.
    </para>
   </listitem>
   <listitem>
    <para>
     Some of them are adaptive &mdash; that means when computers get
     faster, you can tune the algorithm to be slower, without
     introducing incompatibility with existing passwords.
    </para>
   </listitem>
  </orderedlist>

  <table>
   <title>Supported algorithms for <function>crypt()</></title>
   <tgroup cols="5">
    <thead>
     <row>
      <entry>Algorithm</entry>
      <entry>Max password length</entry>
      <entry>Adaptive?</entry>
      <entry>Salt bits</entry>
      <entry>Description</entry>
     </row>
    </thead>
    <tbody>
     <row>
      <entry><literal>bf</></entry>
      <entry>72</entry>
      <entry>yes</entry>
      <entry>128</entry>
      <entry>Blowfish-based, variant 2a</entry>
     </row>
     <row>
      <entry><literal>md5</></entry>
      <entry>unlimited</entry>
      <entry>no</entry>
      <entry>48</entry>
      <entry>MD5-based crypt</entry>
     </row>
     <row>
      <entry><literal>xdes</></entry>
      <entry>8</entry>
      <entry>yes</entry>
      <entry>24</entry>
      <entry>Extended DES</entry>
     </row>
     <row>
      <entry><literal>des</></entry>
      <entry>8</entry>
      <entry>no</entry>
      <entry>12</entry>
      <entry>Original UNIX crypt</entry>
     </row>
    </tbody>
   </tgroup>
  </table>

  <sect3>
   <title><function>crypt()</></title>

   <synopsis>
    crypt(password text, salt text) returns text
   </synopsis>

   <para>
    Calculates a crypt(3)-style hash of <parameter>password</>.
    When storing a new password, you need to use
    <function>gen_salt()</> to generate a new <parameter>salt</> value.
    To check a password, pass the stored hash value as <parameter>salt</>,
    and test whether the result matches the stored value.
   </para>
   <para>
    Example of setting a new password:
   </para>
   <programlisting>
    UPDATE ... SET pswhash = crypt('new password', gen_salt('md5'));
   </programlisting>
   <para>
    Example of authentication:
   </para>
   <programlisting>
    SELECT pswhash = crypt('entered password', pswhash) FROM ... ;
   </programlisting>
   <para>
    This returns <literal>true</> if the entered password is correct.
   </para>
  </sect3>

  <sect3>
   <title><function>gen_salt()</></title>

   <synopsis>
    gen_salt(type text [, iter_count integer ]) returns text
   </synopsis>

   <para>
    Generates a new random salt string for use in <function>crypt()</>.
    The salt string also tells <function>crypt()</> which algorithm to use.
   </para>

   <para>
    The <parameter>type</> parameter specifies the hashing algorithm.
    The accepted types are: <literal>des</literal>, <literal>xdes</literal>,
    <literal>md5</literal> and <literal>bf</literal>.
   </para>

   <para>
    The <parameter>iter_count</> parameter lets the user specify the iteration
    count, for algorithms that have one.
    The higher the count, the more time it takes to hash
    the password and therefore the more time to break it.  Although with
    too high a count the time to calculate a hash may be several years
    &mdash; which is somewhat impractical.  If the <parameter>iter_count</>
    parameter is omitted, the default iteration count is used.
    Allowed values for <parameter>iter_count</> depend on the algorithm:
   </para>

   <table>
    <title>Iteration counts for <function>crypt()</></title>
    <tgroup cols="4">
     <thead>
      <row>
       <entry>Algorithm</entry>
       <entry>Default</entry>
       <entry>Min</entry>
       <entry>Max</entry>
      </row>
     </thead>
     <tbody>
      <row>
       <entry><literal>xdes</></entry>
       <entry>725</entry>
       <entry>1</entry>
       <entry>16777215</entry>
      </row>
      <row>
       <entry><literal>bf</></entry>
       <entry>6</entry>
       <entry>4</entry>
       <entry>31</entry>
      </row>
     </tbody>
    </tgroup>
   </table>

   <para>
    For <literal>xdes</literal> there is an additional limitation that the
    iteration count must be an odd number.
   </para>

   <para>
    To pick an appropriate iteration count, consider that
    the original DES crypt was designed to have the speed of 4 hashes per
    second on the hardware of that time.
    Slower than 4 hashes per second would probably dampen usability.
    Faster than 100 hashes per second is probably too fast.
   </para>

   <para>
    Here is a table that gives an overview of the relative slowness
    of different hashing algorithms.
    The table shows how much time it would take to try all
    combinations of characters in an 8-character password, assuming
    that the password contains either only lowercase letters, or
    upper- and lower-case letters and numbers.
    In the <literal>crypt-bf</literal> entries, the number after a slash is
    the <parameter>iter_count</parameter> parameter of
    <function>gen_salt</function>.
   </para>

   <table>
    <title>Hash algorithm speeds</title>
    <tgroup cols="4">
     <thead>
      <row>
       <entry>Algorithm</entry>
       <entry>Hashes/sec</entry>
       <entry>For <literal>[a-z]</></entry>
       <entry>For <literal>[A-Za-z0-9]</></entry>
      </row>
     </thead>
     <tbody>
      <row>
       <entry><literal>crypt-bf/8</></entry>
       <entry>28</entry>
       <entry>246 years</entry>
       <entry>251322 years</entry>
      </row>
      <row>
       <entry><literal>crypt-bf/7</></entry>
       <entry>57</entry>
       <entry>121 years</entry>
       <entry>123457 years</entry>
      </row>
      <row>
       <entry><literal>crypt-bf/6</></entry>
       <entry>112</entry>
       <entry>62 years</entry>
       <entry>62831 years</entry>
      </row>
      <row>
       <entry><literal>crypt-bf/5</></entry>
       <entry>211</entry>
       <entry>33 years</entry>
       <entry>33351 years</entry>
      </row>
      <row>
       <entry><literal>crypt-md5</></entry>
       <entry>2681</entry>
       <entry>2.6 years</entry>
       <entry>2625 years</entry>
      </row>
      <row>
       <entry><literal>crypt-des</></entry>
       <entry>362837</entry>
       <entry>7 days</entry>
       <entry>19 years</entry>
      </row>
      <row>
       <entry><literal>sha1</></entry>
       <entry>590223</entry>
       <entry>4 days</entry>
       <entry>12 years</entry>
      </row>
      <row>
       <entry><literal>md5</></entry>
       <entry>2345086</entry>
       <entry>1 day</entry>
       <entry>3 years</entry>
      </row>
     </tbody>
    </tgroup>
   </table>

   <para>
    Notes:
   </para>

   <itemizedlist>
    <listitem>
     <para>
     The machine used is a 1.5GHz Pentium 4.
     </para>
    </listitem>
    <listitem>
     <para>
      <literal>crypt-des</> and <literal>crypt-md5</> algorithm numbers are
      taken from John the Ripper v1.6.38 <literal>-test</> output.
     </para>
    </listitem>
    <listitem>
     <para>
      <literal>md5</> numbers are from mdcrack 1.2.
     </para>
    </listitem>
    <listitem>
     <para>
      <literal>sha1</> numbers are from lcrack-20031130-beta.
     </para>
    </listitem>
    <listitem>
     <para>
      <literal>crypt-bf</literal> numbers are taken using a simple program that
      loops over 1000 8-character passwords.  That way I can show the speed
      with different numbers of iterations.  For reference: <literal>john
      -test</literal> shows 213 loops/sec for <literal>crypt-bf/5</>.
      (The very small
      difference in results is in accordance with the fact that the
      <literal>crypt-bf</literal> implementation in <filename>pgcrypto</>
      is the same one used in John the Ripper.)
     </para>
    </listitem>
   </itemizedlist>

   <para>
    Note that <quote>try all combinations</quote> is not a realistic exercise.
    Usually password cracking is done with the help of dictionaries, which
    contain both regular words and various mutations of them.  So, even
    somewhat word-like passwords could be cracked much faster than the above
    numbers suggest, while a 6-character non-word-like password may escape
    cracking.  Or not.
   </para>
  </sect3>
 </sect2>

 <sect2>
  <title>PGP encryption functions</title>

  <para>
   The functions here implement the encryption part of the OpenPGP (RFC 4880)
   standard.  Supported are both symmetric-key and public-key encryption.
  </para>

  <para>
   An encrypted PGP message consists of 2 parts, or <firstterm>packets</>:
  </para>
  <itemizedlist>
   <listitem>
    <para>
     Packet containing a session key &mdash; either symmetric-key or public-key
     encrypted.
    </para>
   </listitem>
   <listitem>
    <para>
     Packet containing data encrypted with the session key.
    </para>
   </listitem>
  </itemizedlist>

  <para>
   When encrypting with a symmetric key (i.e., a password):
  </para>
  <orderedlist>
   <listitem>
    <para>
     The given password is hashed using a String2Key (S2K) algorithm.  This is
     rather similar to <function>crypt()</> algorithms &mdash; purposefully
     slow and with random salt &mdash; but it produces a full-length binary
     key.
    </para>
   </listitem>
   <listitem>
    <para>
     If a separate session key is requested, a new random key will be
     generated.  Otherwise the S2K key will be used directly as the session
     key.
    </para>
   </listitem>
   <listitem>
    <para>
     If the S2K key is to be used directly, then only S2K settings will be put
     into the session key packet.  Otherwise the session key will be encrypted
     with the S2K key and put into the session key packet.
    </para>
   </listitem>
  </orderedlist>

  <para>
   When encrypting with a public key:
  </para>
  <orderedlist>
   <listitem>
    <para>
     A new random session key is generated.
    </para>
   </listitem>
   <listitem>
    <para>
     It is encrypted using the public key and put into the session key packet.
    </para>
   </listitem>
  </orderedlist>

  <para>
   In either case the data to be encrypted is processed as follows:
  </para>
  <orderedlist>
   <listitem>
    <para>
     Optional data-manipulation: compression, conversion to UTF-8,
     and/or conversion of line-endings.
    </para>
   </listitem>
   <listitem>
    <para>
     The data is prefixed with a block of random bytes.  This is equivalent
     to using a random IV.
    </para>
   </listitem>
   <listitem>
    <para>
     An SHA1 hash of the random prefix and data is appended.
    </para>
   </listitem>
   <listitem>
    <para>
     All this is encrypted with the session key and placed in the data packet.
    </para>
   </listitem>
  </orderedlist>

  <sect3>
   <title><function>pgp_sym_encrypt()</function></title>

   <synopsis>
    pgp_sym_encrypt(data text, psw text [, options text ]) returns bytea
    pgp_sym_encrypt_bytea(data bytea, psw text [, options text ]) returns bytea
   </synopsis>
   <para>
    Encrypt <parameter>data</> with a symmetric PGP key <parameter>psw</>.
    The <parameter>options</> parameter can contain option settings,
    as described below.
   </para>
  </sect3>

  <sect3>
   <title><function>pgp_sym_decrypt()</function></title>

   <synopsis>
    pgp_sym_decrypt(msg bytea, psw text [, options text ]) returns text
    pgp_sym_decrypt_bytea(msg bytea, psw text [, options text ]) returns bytea
   </synopsis>
   <para>
    Decrypt a symmetric-key-encrypted PGP message.
   </para>
   <para>
    Decrypting bytea data with <function>pgp_sym_decrypt</> is disallowed.
    This is to avoid outputting invalid character data.  Decrypting
    originally textual data with <function>pgp_sym_decrypt_bytea</> is fine.
   </para>
   <para>
    The <parameter>options</> parameter can contain option settings,
    as described below.
   </para>
  </sect3>

  <sect3>
   <title><function>pgp_pub_encrypt()</function></title>

   <synopsis>
    pgp_pub_encrypt(data text, key bytea [, options text ]) returns bytea
    pgp_pub_encrypt_bytea(data bytea, key bytea [, options text ]) returns bytea
   </synopsis>
   <para>
    Encrypt <parameter>data</> with a public PGP key <parameter>key</>.
    Giving this function a secret key will produce a error.
   </para>
   <para>
    The <parameter>options</> parameter can contain option settings,
    as described below.
   </para>
  </sect3>

  <sect3>
   <title><function>pgp_pub_decrypt()</function></title>

   <synopsis>
    pgp_pub_decrypt(msg bytea, key bytea [, psw text [, options text ]]) returns text
    pgp_pub_decrypt_bytea(msg bytea, key bytea [, psw text [, options text ]]) returns bytea
   </synopsis>
   <para>
    Decrypt a public-key-encrypted message.  <parameter>key</> must be the
    secret key corresponding to the public key that was used to encrypt.
    If the secret key is password-protected, you must give the password in
    <parameter>psw</>.  If there is no password, but you want to specify
    options, you need to give an empty password.
   </para>
   <para>
    Decrypting bytea data with <function>pgp_pub_decrypt</> is disallowed.
    This is to avoid outputting invalid character data.  Decrypting
    originally textual data with <function>pgp_pub_decrypt_bytea</> is fine.
   </para>
   <para>
    The <parameter>options</> parameter can contain option settings,
    as described below.
   </para>
  </sect3>

  <sect3>
   <title><function>pgp_key_id()</function></title>

   <synopsis>
    pgp_key_id(bytea) returns text
   </synopsis>
   <para>
    <function>pgp_key_id</> extracts the key ID of a PGP public or secret key.
    Or it gives the key ID that was used for encrypting the data, if given
    an encrypted message.
   </para>
   <para>
    It can return 2 special key IDs:
   </para>
   <itemizedlist>
    <listitem>
     <para>
      <literal>SYMKEY</>
     </para>
     <para>
      The message is encrypted with a symmetric key.
     </para>
    </listitem>
    <listitem>
     <para>
      <literal>ANYKEY</>
     </para>
     <para>
      The message is public-key encrypted, but the key ID has been removed.
      That means you will need to try all your secret keys on it to see
      which one decrypts it.  <filename>pgcrypto</> itself does not produce
      such messages.
     </para>
    </listitem>
   </itemizedlist>
   <para>
    Note that different keys may have the same ID.   This is rare but a normal
    event. The client application should then try to decrypt with each one,
    to see which fits &mdash; like handling <literal>ANYKEY</>.
   </para>
  </sect3>

  <sect3>
   <title><function>armor()</function>, <function>dearmor()</function></title>

   <synopsis>
    armor(data bytea) returns text
    dearmor(data text) returns bytea
   </synopsis>
   <para>
    These functions wrap/unwrap binary data into PGP Ascii Armor format,
    which is basically Base64 with CRC and additional formatting.
   </para>
  </sect3>

  <sect3>
   <title>Options for PGP functions</title>

   <para>
    Options are named to be similar to GnuPG.  An option's value should be
    given after an equal sign; separate options from each other with commas.
    For example:
   </para>
   <programlisting>
    pgp_sym_encrypt(data, psw, 'compress-algo=1, cipher-algo=aes256')
   </programlisting>

   <para>
    All of the options except <literal>convert-crlf</literal> apply only to
    encrypt functions.  Decrypt functions get the parameters from the PGP
    data.
   </para>

   <para>
    The most interesting options are probably
    <literal>compress-algo</literal> and <literal>unicode-mode</literal>.
    The rest should have reasonable defaults.
   </para>

  <sect4>
   <title>cipher-algo</title>

   <para>
    Which cipher algorithm to use.
   </para>
   <programlisting>
    Values: bf, aes128, aes192, aes256 (OpenSSL-only: <literal>3des</literal>, <literal>cast5</literal>)
    Default: aes128
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>compress-algo</title>

   <para>
    Which compression algorithm to use.  Only available if
    <productname>PostgreSQL</productname> was built with zlib.
   </para>
   <programlisting>
    Values:
      0 - no compression
      1 - ZIP compression
      2 - ZLIB compression (= ZIP plus meta-data and block CRCs)
    Default: 0
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>compress-level</title>

   <para>
    How much to compress.  Higher levels compress smaller but are slower.
    0 disables compression.
   </para>
   <programlisting>
    Values: 0, 1-9
    Default: 6
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>convert-crlf</title>

   <para>
    Whether to convert <literal>\n</literal> into <literal>\r\n</literal> when
    encrypting and <literal>\r\n</literal> to <literal>\n</literal> when
    decrypting.  RFC 4880 specifies that text data should be stored using
    <literal>\r\n</literal> line-feeds.  Use this to get fully RFC-compliant
    behavior.
   </para>
   <programlisting>
    Values: 0, 1
    Default: 0
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt, pgp_sym_decrypt, pgp_pub_decrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>disable-mdc</title>

   <para>
    Do not protect data with SHA-1.  The only good reason to use this
    option is to achieve compatibility with ancient PGP products, predating
    the addition of SHA-1 protected packets to RFC 4880.
    Recent gnupg.org and pgp.com software supports it fine.
   </para>
   <programlisting>
    Values: 0, 1
    Default: 0
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>enable-session-key</title>

   <para>
    Use separate session key.  Public-key encryption always uses a separate
    session key; this is for symmetric-key encryption, which by default
    uses the S2K key directly.
   </para>
   <programlisting>
    Values: 0, 1
    Default: 0
    Applies to: pgp_sym_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>s2k-mode</title>

   <para>
    Which S2K algorithm to use.
   </para>
   <programlisting>
    Values:
      0 - Without salt.  Dangerous!
      1 - With salt but with fixed iteration count.
      3 - Variable iteration count.
    Default: 3
    Applies to: pgp_sym_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>s2k-digest-algo</title>

   <para>
    Which digest algorithm to use in S2K calculation.
   </para>
   <programlisting>
    Values: md5, sha1
    Default: sha1
    Applies to: pgp_sym_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>s2k-cipher-algo</title>

   <para>
    Which cipher to use for encrypting separate session key.
   </para>
   <programlisting>
    Values: bf, aes, aes128, aes192, aes256
    Default: use cipher-algo
    Applies to: pgp_sym_encrypt
   </programlisting>
  </sect4>

  <sect4>
   <title>unicode-mode</title>

   <para>
    Whether to convert textual data from database internal encoding to
    UTF-8 and back.  If your database already is UTF-8, no conversion will
    be done, but the message will be tagged as UTF-8.  Without this option
    it will not be.
   </para>
   <programlisting>
    Values: 0, 1
    Default: 0
    Applies to: pgp_sym_encrypt, pgp_pub_encrypt
   </programlisting>
  </sect4>
  </sect3>

 <sect3>
  <title>Generating PGP keys with GnuPG</title>

  <para>
   To generate a new key:
  </para>
  <programlisting>
   gpg --gen-key
  </programlisting>
  <para>
   The preferred key type is <quote>DSA and Elgamal</>.
  </para>
  <para>
   For RSA encryption you must create either DSA or RSA sign-only key
   as master and then add an RSA encryption subkey with
   <literal>gpg --edit-key</literal>.
  </para>
  <para>
   To list keys:
  </para>
  <programlisting>
   gpg --list-secret-keys
  </programlisting>
  <para>
   To export a public key in ascii-armor format:
  </para>
  <programlisting>
   gpg -a --export KEYID > public.key
  </programlisting>
  <para>
   To export a secret key in ascii-armor format:
  </para>
  <programlisting>
   gpg -a --export-secret-keys KEYID > secret.key
  </programlisting>
  <para>
   You need to use <function>dearmor()</> on these keys before giving them to
   the PGP functions.  Or if you can handle binary data, you can drop
   <literal>-a</literal> from the command.
  </para>
  <para>
   For more details see <literal>man gpg</literal>,
   <ulink url="http://www.gnupg.org/gph/en/manual.html">The GNU
   Privacy Handbook</ulink> and other documentation on
   <ulink url="http://www.gnupg.org"></ulink>.
  </para>
 </sect3>

 <sect3>
  <title>Limitations of PGP code</title>

  <itemizedlist>
   <listitem>
    <para>
    No support for signing.  That also means that it is not checked
    whether the encryption subkey belongs to the master key.
    </para>
   </listitem>
   <listitem>
    <para>
    No support for encryption key as master key.  As such practice
    is generally discouraged, this should not be a problem.
    </para>
   </listitem>
   <listitem>
    <para>
    No support for several subkeys.  This may seem like a problem, as this
    is common practice.  On the other hand, you should not use your regular
    GPG/PGP keys with <filename>pgcrypto</>, but create new ones,
    as the usage scenario is rather different.
    </para>
   </listitem>
  </itemizedlist>
  </sect3>
 </sect2>

 <sect2>
  <title>Raw encryption functions</title>

  <para>
   These functions only run a cipher over data; they don't have any advanced
   features of PGP encryption.  Therefore they have some major problems:
  </para>
  <orderedlist>
   <listitem>
    <para>
    They use user key directly as cipher key.
    </para>
   </listitem>
   <listitem>
    <para>
    They don't provide any integrity checking, to see
    if the encrypted data was modified.
    </para>
   </listitem>
   <listitem>
    <para>
    They expect that users manage all encryption parameters
    themselves, even IV.
    </para>
   </listitem>
   <listitem>
    <para>
    They don't handle text.
    </para>
   </listitem>
  </orderedlist>
  <para>
   So, with the introduction of PGP encryption, usage of raw
   encryption functions is discouraged.
  </para>

  <synopsis>
    encrypt(data bytea, key bytea, type text) returns bytea
    decrypt(data bytea, key bytea, type text) returns bytea

    encrypt_iv(data bytea, key bytea, iv bytea, type text) returns bytea
    decrypt_iv(data bytea, key bytea, iv bytea, type text) returns bytea
  </synopsis>

  <para>
   Encrypt/decrypt data using the cipher method specified by
   <parameter>type</parameter>.  The syntax of the
   <parameter>type</parameter> string is:
  </para>

  <synopsis>
   <replaceable>algorithm</> <optional> <literal>-</> <replaceable>mode</> </optional> <optional> <literal>/pad:</> <replaceable>padding</> </optional>
  </synopsis>

  <para>
   where <replaceable>algorithm</> is one of:
  </para>
  <itemizedlist>
   <listitem><para><literal>bf</literal> &mdash; Blowfish</para></listitem>
   <listitem><para><literal>aes</literal> &mdash; AES (Rijndael-128)</para></listitem>
  </itemizedlist>
  <para>
   and <replaceable>mode</> is one of:
  </para>
  <itemizedlist>
   <listitem>
    <para>
    <literal>cbc</literal> &mdash; next block depends on previous (default)
    </para>
   </listitem>
   <listitem>
    <para>
    <literal>ecb</literal> &mdash; each block is encrypted separately (for
    testing only)
    </para>
   </listitem>
  </itemizedlist>
  <para>
   and <replaceable>padding</> is one of:
  </para>
  <itemizedlist>
   <listitem>
    <para>
    <literal>pkcs</literal> &mdash; data may be any length (default)
    </para>
   </listitem>
   <listitem>
    <para>
    <literal>none</literal> &mdash; data must be multiple of cipher block size
    </para>
   </listitem>
  </itemizedlist>
  <para>
   So, for example, these are equivalent:
  </para>
  <programlisting>
   encrypt(data, 'fooz', 'bf')
   encrypt(data, 'fooz', 'bf-cbc/pad:pkcs')
  </programlisting>
  <para>
   In <function>encrypt_iv</> and <function>decrypt_iv</>, the
   <parameter>iv</> parameter is the initial value for the CBC mode;
   it is ignored for ECB.
   It is clipped or padded with zeroes if not exactly block size.
   It defaults to all zeroes in the functions without this parameter.
  </para>
 </sect2>

 <sect2>
  <title>Random-data functions</title>

  <synopsis>
   gen_random_bytes(count integer) returns bytea
  </synopsis>
  <para>
   Returns <parameter>count</> cryptographically strong random bytes.
   At most 1024 bytes can be extracted at a time.  This is to avoid
   draining the randomness generator pool.
  </para>
 </sect2>

 <sect2>
  <title>Notes</title>

  <sect3>
   <title>Configuration</title>

   <para>
    <filename>pgcrypto</> configures itself according to the findings of the
    main PostgreSQL <literal>configure</literal> script.  The options that
    affect it are <literal>--with-zlib</literal> and
    <literal>--with-openssl</literal>.
   </para>

   <para>
    When compiled with zlib, PGP encryption functions are able to
    compress data before encrypting.
   </para>

   <para>
    When compiled with OpenSSL, there will be more algorithms available.
    Also public-key encryption functions will be faster as OpenSSL
    has more optimized BIGNUM functions.
   </para>

   <table id="pgcrypto-with-without-openssl">
    <title>Summary of functionality with and without OpenSSL</title>
    <tgroup cols="3">
     <thead>
      <row>
       <entry>Functionality</entry>
       <entry>Built-in</entry>
       <entry>With OpenSSL</entry>
      </row>
     </thead>
     <tbody>
      <row>
       <entry>MD5</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>SHA1</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>SHA224/256/384/512</entry>
       <entry>yes</entry>
       <entry>yes (Note 1)</entry>
      </row>
      <row>
       <entry>Other digest algorithms</entry>
       <entry>no</entry>
       <entry>yes (Note 2)</entry>
      </row>
      <row>
       <entry>Blowfish</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>AES</entry>
       <entry>yes</entry>
       <entry>yes (Note 3)</entry>
      </row>
      <row>
       <entry>DES/3DES/CAST5</entry>
       <entry>no</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>Raw encryption</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>PGP Symmetric encryption</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
      <row>
       <entry>PGP Public-Key encryption</entry>
       <entry>yes</entry>
       <entry>yes</entry>
      </row>
     </tbody>
    </tgroup>
   </table>

   <para>
    Notes:
   </para>

   <orderedlist>
    <listitem>
     <para>
      SHA2 algorithms were added to OpenSSL in version 0.9.8.  For
      older versions, <filename>pgcrypto</> will use built-in code.
     </para>
    </listitem>
    <listitem>
     <para>
      Any digest algorithm OpenSSL supports is automatically picked up.
      This is not possible with ciphers, which need to be supported
      explicitly.
     </para>
    </listitem>
    <listitem>
     <para>
      AES is included in OpenSSL since version 0.9.7.  For
      older versions, <filename>pgcrypto</> will use built-in code.
     </para>
    </listitem>
   </orderedlist>
  </sect3>

  <sect3>
   <title>NULL handling</title>

   <para>
    As is standard in SQL, all functions return NULL, if any of the arguments
    are NULL.  This may create security risks on careless usage.
   </para>
  </sect3>

  <sect3>
   <title>Security limitations</title>

   <para>
    All <filename>pgcrypto</> functions run inside the database server.
    That means that all
    the data and passwords move between <filename>pgcrypto</> and client
    applications in clear text.  Thus you must:
   </para>

   <orderedlist>
    <listitem>
     <para>Connect locally or use SSL connections.</para>
    </listitem>
    <listitem>
     <para>Trust both system and database administrator.</para>
    </listitem>
   </orderedlist>

   <para>
    If you cannot, then better do crypto inside client application.
   </para>
  </sect3>

  <sect3>
   <title>Useful reading</title>

   <itemizedlist>
    <listitem>
     <para><ulink url="http://www.gnupg.org/gph/en/manual.html"></ulink></para>
     <para>The GNU Privacy Handbook.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://www.openwall.com/crypt/"></ulink></para>
     <para>Describes the crypt-blowfish algorithm.</para>
    </listitem>
    <listitem>
     <para>
      <ulink url="http://www.stack.nl/~galactus/remailers/passphrase-faq.html"></ulink>
     </para>
     <para>How to choose a good password.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://world.std.com/~reinhold/diceware.html"></ulink></para>
     <para>Interesting idea for picking passwords.</para>
    </listitem>
    <listitem>
     <para>
      <ulink url="http://www.interhack.net/people/cmcurtin/snake-oil-faq.html"></ulink>
     </para>
     <para>Describes good and bad cryptography.</para>
    </listitem>
   </itemizedlist>
  </sect3>

  <sect3>
   <title>Technical references</title>

   <itemizedlist>
    <listitem>
     <para><ulink url="http://www.ietf.org/rfc/rfc4880.txt"></ulink></para>
     <para>OpenPGP message format.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://www.ietf.org/rfc/rfc1321.txt"></ulink></para>
     <para>The MD5 Message-Digest Algorithm.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://www.ietf.org/rfc/rfc2104.txt"></ulink></para>
     <para>HMAC: Keyed-Hashing for Message Authentication.</para>
    </listitem>
    <listitem>
     <para>
      <ulink url="http://www.usenix.org/events/usenix99/provos.html"></ulink>
     </para>
     <para>Comparison of crypt-des, crypt-md5 and bcrypt algorithms.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://csrc.nist.gov/cryptval/des.htm"></ulink></para>
     <para>Standards for DES, 3DES and AES.</para>
    </listitem>
    <listitem>
     <para>
      <ulink url="http://en.wikipedia.org/wiki/Fortuna_(PRNG)"></ulink>
     </para>
     <para>Description of Fortuna CSPRNG.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://jlcooke.ca/random/"></ulink></para>
     <para>Jean-Luc Cooke Fortuna-based /dev/random driver for Linux.</para>
    </listitem>
    <listitem>
     <para><ulink url="http://www.cs.ut.ee/~helger/crypto/"></ulink></para>
     <para>Collection of cryptology pointers.</para>
    </listitem>
   </itemizedlist>
  </sect3>
 </sect2>

 <sect2>
  <title>Author</title>

  <para>
   Marko Kreen <email>markokr@gmail.com</email>
  </para>

  <para>
   <filename>pgcrypto</filename> uses code from the following sources:
  </para>

  <table>
   <title>Credits</title>
   <tgroup cols="3">
    <thead>
     <row>
      <entry>Algorithm</entry>
      <entry>Author</entry>
      <entry>Source origin</entry>
     </row>
    </thead>
    <tbody>
     <row>
      <entry>DES crypt</entry>
      <entry>David Burren and others</entry>
      <entry>FreeBSD libcrypt</entry>
     </row>
     <row>
      <entry>MD5 crypt</entry>
      <entry>Poul-Henning Kamp</entry>
      <entry>FreeBSD libcrypt</entry>
     </row>
     <row>
      <entry>Blowfish crypt</entry>
      <entry>Solar Designer</entry>
      <entry>www.openwall.com</entry>
     </row>
     <row>
      <entry>Blowfish cipher</entry>
      <entry>Simon Tatham</entry>
      <entry>PuTTY</entry>
     </row>
     <row>
      <entry>Rijndael cipher</entry>
      <entry>Brian Gladman</entry>
      <entry>OpenBSD sys/crypto</entry>
     </row>
     <row>
      <entry>MD5 and SHA1</entry>
      <entry>WIDE Project</entry>
      <entry>KAME kame/sys/crypto</entry>
     </row>
     <row>
      <entry>SHA256/384/512 </entry>
      <entry>Aaron D. Gifford</entry>
      <entry>OpenBSD sys/crypto</entry>
     </row>
     <row>
      <entry>BIGNUM math</entry>
      <entry>Michael J. Fromberger</entry>
      <entry>dartmouth.edu/~sting/sw/imath</entry>
     </row>
    </tbody>
   </tgroup>
  </table>
 </sect2>

</sect1>