/**
* Generate a random string.
*
* Although microoptimizations are generally discouraged as they impair readability this function is ripe with
* microoptimizations because this function has the potential of being called a huge number of times.
* eg. for RSA key generation.
*
* @param int $length
* @return string
*/
static function string($length)
{
if (version_compare(PHP_VERSION, '7.0.0', '>=')) {
try {
return \random_bytes($length);
} catch (\Throwable $e) {
// If a sufficient source of randomness is unavailable, random_bytes() will throw an
// object that implements the Throwable interface (Exception, TypeError, Error).
// We don't actually need to do anything here. The string() method should just continue
// as normal. Note, however, that if we don't have a sufficient source of randomness for
// random_bytes(), most of the other calls here will fail too, so we'll end up using
// the PHP implementation.
}
}
if (strtoupper(substr(PHP_OS, 0, 3)) === 'WIN') {
// method 1. prior to PHP 5.3 this would call rand() on windows hence the function_exists('class_alias') call.
// ie. class_alias is a function that was introduced in PHP 5.3
if (extension_loaded('mcrypt') && function_exists('class_alias')) {
return mcrypt_create_iv($length);
}
// method 2. openssl_random_pseudo_bytes was introduced in PHP 5.3.0 but prior to PHP 5.3.4 there was,
// to quote <http://php.net/ChangeLog-5.php#5.3.4>, "possible blocking behavior". as of 5.3.4
// openssl_random_pseudo_bytes and mcrypt_create_iv do the exact same thing on Windows. ie. they both
// call php_win32_get_random_bytes():
//
// https://github.com/php/php-src/blob/7014a0eb6d1611151a286c0ff4f2238f92c120d6/ext/openssl/openssl.c#L5008
// https://github.com/php/php-src/blob/7014a0eb6d1611151a286c0ff4f2238f92c120d6/ext/mcrypt/mcrypt.c#L1392
//
// php_win32_get_random_bytes() is defined thusly:
//
// https://github.com/php/php-src/blob/7014a0eb6d1611151a286c0ff4f2238f92c120d6/win32/winutil.c#L80
//
// we're calling it, all the same, in the off chance that the mcrypt extension is not available
if (extension_loaded('openssl') && version_compare(PHP_VERSION, '5.3.4', '>=')) {
return openssl_random_pseudo_bytes($length);
}
} else {
// method 1. the fastest
if (extension_loaded('openssl')) {
return openssl_random_pseudo_bytes($length);
}
// method 2
static $fp = true;
if ($fp === true) {
// warning's will be output unles the error suppression operator is used. errors such as
// "open_basedir restriction in effect", "Permission denied", "No such file or directory", etc.
$fp = @fopen('/dev/urandom', 'rb');
}
if ($fp !== true && $fp !== false) {
// surprisingly faster than !is_bool() or is_resource()
return fread($fp, $length);
}
// method 3. pretty much does the same thing as method 2 per the following url:
// https://github.com/php/php-src/blob/7014a0eb6d1611151a286c0ff4f2238f92c120d6/ext/mcrypt/mcrypt.c#L1391
// surprisingly slower than method 2. maybe that's because mcrypt_create_iv does a bunch of error checking that we're
// not doing. regardless, this'll only be called if this PHP script couldn't open /dev/urandom due to open_basedir
// restrictions or some such
if (extension_loaded('mcrypt')) {
return mcrypt_create_iv($length, MCRYPT_DEV_URANDOM);
}
}
// at this point we have no choice but to use a pure-PHP CSPRNG
// cascade entropy across multiple PHP instances by fixing the session and collecting all
// environmental variables, including the previous session data and the current session
// data.
//
// mt_rand seeds itself by looking at the PID and the time, both of which are (relatively)
// easy to guess at. linux uses mouse clicks, keyboard timings, etc, as entropy sources, but
// PHP isn't low level to be able to use those as sources and on a web server there's not likely
// going to be a ton of keyboard or mouse action. web servers do have one thing that we can use
// however, a ton of people visiting the website. obviously you don't want to base your seeding
// soley on parameters a potential attacker sends but (1) not everything in $_SERVER is controlled
// by the user and (2) this isn't just looking at the data sent by the current user - it's based
// on the data sent by all users. one user requests the page and a hash of their info is saved.
// another user visits the page and the serialization of their data is utilized along with the
// server envirnment stuff and a hash of the previous http request data (which itself utilizes
// a hash of the session data before that). certainly an attacker should be assumed to have
// full control over his own http requests. he, however, is not going to have control over
// everyone's http requests.
static $crypto = false, $v;
if ($crypto === false) {
// save old session data
$old_session_id = session_id();
$old_use_cookies = ini_get('session.use_cookies');
$old_session_cache_limiter = session_cache_limiter();
$_OLD_SESSION = isset($_SESSION) ? $_SESSION : false;
if ($old_session_id != '') {
session_write_close();
}
session_id(1);
ini_set('session.use_cookies', 0);
session_cache_limiter('');
session_start();
$v = $seed = $_SESSION['seed'] = pack('H*', sha1(serialize($_SERVER) . serialize($_POST) . serialize($_GET) . serialize($_COOKIE) . serialize($GLOBALS) . serialize($_SESSION) . serialize($_OLD_SESSION)));
if (!isset($_SESSION['count'])) {
$_SESSION['count'] = 0;
}
$_SESSION['count']++;
session_write_close();
// restore old session data
if ($old_session_id != '') {
session_id($old_session_id);
session_start();
ini_set('session.use_cookies', $old_use_cookies);
session_cache_limiter($old_session_cache_limiter);
} else {
if ($_OLD_SESSION !== false) {
$_SESSION = $_OLD_SESSION;
unset($_OLD_SESSION);
} else {
unset($_SESSION);
}
}
// in SSH2 a shared secret and an exchange hash are generated through the key exchange process.
// the IV client to server is the hash of that "nonce" with the letter A and for the encryption key it's the letter C.
// if the hash doesn't produce enough a key or an IV that's long enough concat successive hashes of the
// original hash and the current hash. we'll be emulating that. for more info see the following URL:
//
// http://tools.ietf.org/html/rfc4253#section-7.2
//
// see the is_string($crypto) part for an example of how to expand the keys
$key = pack('H*', sha1($seed . 'A'));
$iv = pack('H*', sha1($seed . 'C'));
// ciphers are used as per the nist.gov link below. also, see this link:
//
// http://en.wikipedia.org/wiki/Cryptographically_secure_pseudorandom_number_generator#Designs_based_on_cryptographic_primitives
switch (true) {
case class_exists('\\phpseclib\\Crypt\\AES'):
$crypto = new AES(Base::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\Twofish'):
$crypto = new Twofish(Base::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\Blowfish'):
$crypto = new Blowfish(Base::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\TripleDES'):
$crypto = new TripleDES(Base::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\DES'):
$crypto = new DES(Base::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\RC4'):
$crypto = new RC4();
break;
default:
user_error(__CLASS__ . ' requires at least one symmetric cipher be loaded');
return false;
}
$crypto->setKey($key);
$crypto->setIV($iv);
$crypto->enableContinuousBuffer();
}
//return $crypto->encrypt(str_repeat("\0", $length));
// the following is based off of ANSI X9.31:
//
// http://csrc.nist.gov/groups/STM/cavp/documents/rng/931rngext.pdf
//
// OpenSSL uses that same standard for it's random numbers:
//
// http://www.opensource.apple.com/source/OpenSSL/OpenSSL-38/openssl/fips-1.0/rand/fips_rand.c
// (do a search for "ANS X9.31 A.2.4")
$result = '';
while (strlen($result) < $length) {
$i = $crypto->encrypt(microtime());
// strlen(microtime()) == 21
$r = $crypto->encrypt($i ^ $v);
// strlen($v) == 20
$v = $crypto->encrypt($r ^ $i);
// strlen($r) == 20
$result .= $r;
}
return substr($result, 0, $length);
}
/**
* Generate a random string.
*
* Although microoptimizations are generally discouraged as they impair readability this function is ripe with
* microoptimizations because this function has the potential of being called a huge number of times.
* eg. for RSA key generation.
*
* @param int $length
* @throws \RuntimeException if a symmetric cipher is needed but not loaded
* @return string
*/
static function string($length)
{
try {
return \random_bytes($length);
} catch (\Exception $e) {
// random_compat will throw an Exception, which in PHP 5 does not implement Throwable
} catch (\Throwable $e) {
// If a sufficient source of randomness is unavailable, random_bytes() will throw an
// object that implements the Throwable interface (Exception, TypeError, Error).
// We don't actually need to do anything here. The string() method should just continue
// as normal. Note, however, that if we don't have a sufficient source of randomness for
// random_bytes(), most of the other calls here will fail too, so we'll end up using
// the PHP implementation.
}
// at this point we have no choice but to use a pure-PHP CSPRNG
// cascade entropy across multiple PHP instances by fixing the session and collecting all
// environmental variables, including the previous session data and the current session
// data.
//
// mt_rand seeds itself by looking at the PID and the time, both of which are (relatively)
// easy to guess at. linux uses mouse clicks, keyboard timings, etc, as entropy sources, but
// PHP isn't low level to be able to use those as sources and on a web server there's not likely
// going to be a ton of keyboard or mouse action. web servers do have one thing that we can use
// however, a ton of people visiting the website. obviously you don't want to base your seeding
// soley on parameters a potential attacker sends but (1) not everything in $_SERVER is controlled
// by the user and (2) this isn't just looking at the data sent by the current user - it's based
// on the data sent by all users. one user requests the page and a hash of their info is saved.
// another user visits the page and the serialization of their data is utilized along with the
// server envirnment stuff and a hash of the previous http request data (which itself utilizes
// a hash of the session data before that). certainly an attacker should be assumed to have
// full control over his own http requests. he, however, is not going to have control over
// everyone's http requests.
static $crypto = false, $v;
if ($crypto === false) {
// save old session data
$old_session_id = session_id();
$old_use_cookies = ini_get('session.use_cookies');
$old_session_cache_limiter = session_cache_limiter();
$_OLD_SESSION = isset($_SESSION) ? $_SESSION : false;
if ($old_session_id != '') {
session_write_close();
}
session_id(1);
ini_set('session.use_cookies', 0);
session_cache_limiter('');
session_start();
$v = (isset($_SERVER) ? self::safe_serialize($_SERVER) : '') . (isset($_POST) ? self::safe_serialize($_POST) : '') . (isset($_GET) ? self::safe_serialize($_GET) : '') . (isset($_COOKIE) ? self::safe_serialize($_COOKIE) : '') . self::safe_serialize($GLOBALS) . self::safe_serialize($_SESSION) . self::safe_serialize($_OLD_SESSION);
$v = $seed = $_SESSION['seed'] = sha1($v, true);
if (!isset($_SESSION['count'])) {
$_SESSION['count'] = 0;
}
$_SESSION['count']++;
session_write_close();
// restore old session data
if ($old_session_id != '') {
session_id($old_session_id);
session_start();
ini_set('session.use_cookies', $old_use_cookies);
session_cache_limiter($old_session_cache_limiter);
} else {
if ($_OLD_SESSION !== false) {
$_SESSION = $_OLD_SESSION;
unset($_OLD_SESSION);
} else {
unset($_SESSION);
}
}
// in SSH2 a shared secret and an exchange hash are generated through the key exchange process.
// the IV client to server is the hash of that "nonce" with the letter A and for the encryption key it's the letter C.
// if the hash doesn't produce enough a key or an IV that's long enough concat successive hashes of the
// original hash and the current hash. we'll be emulating that. for more info see the following URL:
//
// http://tools.ietf.org/html/rfc4253#section-7.2
//
// see the is_string($crypto) part for an example of how to expand the keys
$key = sha1($seed . 'A', true);
$iv = sha1($seed . 'C', true);
// ciphers are used as per the nist.gov link below. also, see this link:
//
// http://en.wikipedia.org/wiki/Cryptographically_secure_pseudorandom_number_generator#Designs_based_on_cryptographic_primitives
switch (true) {
case class_exists('\\phpseclib\\Crypt\\AES'):
$crypto = new AES(BlockCipher::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\Twofish'):
$crypto = new Twofish(BlockCipher::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\Blowfish'):
$crypto = new Blowfish(BlockCipher::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\TripleDES'):
$crypto = new TripleDES(BlockCipher::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\DES'):
$crypto = new DES(BlockCipher::MODE_CTR);
break;
case class_exists('\\phpseclib\\Crypt\\RC4'):
$crypto = new RC4();
break;
default:
throw new \RuntimeException(__CLASS__ . ' requires at least one symmetric cipher be loaded');
}
$crypto->setKey(substr($key, 0, $crypto->getKeyLength() >> 3));
$crypto->setIV(substr($iv, 0, $crypto->getBlockLength() >> 3));
$crypto->enableContinuousBuffer();
}
//return $crypto->encrypt(str_repeat("\0", $length));
// the following is based off of ANSI X9.31:
//
// http://csrc.nist.gov/groups/STM/cavp/documents/rng/931rngext.pdf
//
// OpenSSL uses that same standard for it's random numbers:
//
// http://www.opensource.apple.com/source/OpenSSL/OpenSSL-38/openssl/fips-1.0/rand/fips_rand.c
// (do a search for "ANS X9.31 A.2.4")
$result = '';
while (strlen($result) < $length) {
$i = $crypto->encrypt(microtime());
// strlen(microtime()) == 21
$r = $crypto->encrypt($i ^ $v);
// strlen($v) == 20
$v = $crypto->encrypt($r ^ $i);
// strlen($r) == 20
$result .= $r;
}
return substr($result, 0, $length);
}