(PHP 4, PHP 5, PHP 7, PHP 8)
unserialize — 保存用表現から PHP の値を生成する
unserialize() は、シリアル化された変数を PHP の値に戻す変換(アンシリアライズ)を行います。
allowed_classes
の options
の値にかかわらず、
ユーザーからの入力をそのまま unserialize() に渡してはいけません。
アンシリアライズの時には、オブジェクトのインスタンス生成やオートローディングなどで
コードが実行されることがあり、悪意のあるユーザーがこれを悪用するかもしれないからです。
シリアル化したデータをユーザーに渡す必要がある場合は、安全で標準的なデータ交換フォーマットである
JSON などを使うようにしましょう。
json_decode() および json_encode()
を利用します。
外部に保存されているシリアル化されたデータをアンシリアライズする必要がある場合は、 hash_hmac() を使ったデータの検証を検討しましょう。 他者によるデータの改ざんがないことを確かめるためです。
data
シリアル化された文字列。
もしアンシリアライズする変数がオブジェクトの場合、 オブジェクトが無事再作成された後、PHP は自動的にメンバ関数 __unserialize() または __wakeup() (存在していれば) をコールしようとします。
注意: unserialize_callback_func ディレクティブ
unserialize_callback_func ディレクティブで指定したコールバックは、未定義のクラスをアンシリアライズしようとした場合にコールされます。 コールバックが指定されない場合は、__PHP_Incomplete_Class がインスタンス化されます。
options
unserialize() に連想配列で渡すオプション。
名前 | 型 | 説明 |
---|---|---|
allowed_classes |
array |bool |
受け付けるクラス名の配列を指定します。あらゆるクラスを拒否する場合は
false 、あらゆるクラスを受け付ける場合は true を指定します。
このオプションを指定しているときに、もし
unserialize() が受け付け対象外のクラスのオブジェクトに遭遇すると、
そのオブジェクトを
__PHP_Incomplete_Class のインスタンスに変換します。
このオプションを省略すると、true を指定したのと同じ意味になります。
つまり、PHP はあらゆるクラスのオブジェクトをインスタンス化しようとします。
|
max_depth |
int |
アンシリアライズ処理の間に許される、
データ構造の再帰の深さの最大値を設定します。
これは、スタックオーバーフローを防ぐためのものです。
デフォルトの深さの最大値は 4096 であり、
0 に設定すると、
この制限を無効にすることができます。
|
オブジェクトは、 アンシリアライズを実行するハンドラで Throwable をスローしても構いません。
PHP 8.4.0 以降では、options
の allowed_classes
要素が
クラス名の array でない場合、
unserialize() は TypeError と
ValueError をスローします。
バージョン | 説明 |
---|---|
8.4.0 |
options の allowed_classes 要素が
クラス名の array でない場合、
TypeError と ValueError を
スローするようになりました。
|
8.3.0 |
入力された文字列の末尾に余計なデータがある場合、E_WARNING が発生するようになりました。
|
8.3.0 |
渡された文字列が復元できない場合、
E_WARNING が発生するようになりました。
これより前のバージョンでは、E_NOTICE
が発生していました。
|
7.4.0 |
options に max_depth が追加されました。
これは、アンシリアライズ処理の間に許される、
データ構造の再帰の深さの最大値を設定するものです。
|
7.1.0 |
options の allowed_classes 要素は、
型を厳密に調べるようになりました。
つまり、array または bool 以外の型が与えられると、
unserialize() 関数は false を返し、
E_WARNING レベルの警告を発生させます。
|
例1 unserialize() の例
<?php
// ここで、データベースから $session_data にセッションデータをロード
// するために unserialize() を使用します。
// この例は、serialize() で記述された例を補足するものです。
$conn = odbc_connect("webdb", "php", "chicken");
$stmt = odbc_prepare($conn, "SELECT data FROM sessions WHERE id = ?");
$sqldata = array($_SERVER['PHP_AUTH_USER']);
if (!odbc_execute($stmt, $sqldata) || !odbc_fetch_into($stmt, $tmp)) {
// 実行または取得が失敗した場合、空の配列で初期化します
$session_data = array();
} else {
// tmp[0] にシリアル化されたデータを保持している必要があります。
$session_data = unserialize($tmp[0]);
if (!is_array($session_data)) {
// 何か問題があったため、空の配列で初期化します。
$session_data = array();
}
}
?>
例2 unserialize_callback_func の例
<?php
$serialized_object='O:1:"a":1:{s:5:"value";s:3:"100";}';
ini_set('unserialize_callback_func', 'mycallback'); // 独自のコールバック関数を設定する
function mycallback($classname)
{
// クラスが定義されているファイルをインクルードするだけです。
// どのクラス定義が必要になるのかを $classname で判断します。
var_dump($classname);
}
unserialize($serialized_object);
?>
Just some reminder which may save somebody some time regarding the `$options` array:
Say you want to be on the safe side and not allow any objects to be unserialized... My first thought was doing the following:
<?php
$lol = unserialize($string, false);
// This will generate:
// Warning: unserialize() expects parameter 2 to be array, boolean given
?>
The correct way of doing this is the following:
<?php
$lol = unserialize($string, ['allowed_classes' => false]);
?>
Hope it helps somebody!
Keep in mind that the allowed_classes does not use inheritance, i.e. allowing an interface is not possible and sub-classes won't pass the check. See https://3v4l.org/tdHfl
Just a note - if the serialized string contains a reference to a class that cannot be instantiated (e.g. being abstract) PHP will immediately die with a fatal error. If the unserialize() statement is preceded with a '@' to avoid cluttering the logs with warns or notices there will be absolutely no clue as to why the script stopped working. Cost me a couple of hours...
Here's a simple function to get the class of a serialized string (that is, the type of object that will be returned if it's unserialized):
<?php
function get_serial_class($serial) {
$types = array('s' => 'string', 'a' => 'array', 'b' => 'bool', 'i' => 'int', 'd' => 'float', 'N;' => 'NULL');
$parts = explode(':', $serial, 4);
return isset($types[$parts[0]]) ? $types[$parts[0]] : trim($parts[2], '"');
}
?>
I use this when saving a serialized object to a cookie, to make sure it is the right type when I go to unserialize it.
The type names are the same format/case as you would see if you did a var_dump().
For the people who are getting the error
PHP Notice: unserialize(): Error at offset 191 of 285 bytes in ...
and are getting the data from a database, Make sure that you have the database set the the correct encoding, I had the database set as latin1_swedish_ci and all of the data looked perfect, Infact when i copied it into a online unserialize it worked fine. I changed the collation to utf8mb4_unicode_ci and all worked fine.
Talk on Exploiting PHP7 Unserialize here: https://media.ccc.de/v/33c3-7858-exploiting_php7_unserialize
In the Classes and Objects docs, there is this: In order to be able to unserialize() an object, the class of that object needs to be defined.
Prior to PHP 5.3, this was not an issue. But after PHP 5.3 an object made by SimpleXML_Load_String() cannot be serialized. An attempt to do so will result in a run-time failure, throwing an exception. If you store such an object in $_SESSION, you will get a post-execution error that says this:
Fatal error: Uncaught exception 'Exception' with message 'Serialization of 'SimpleXMLElement' is not allowed' in [no active file]:0 Stack trace: #0 {main} thrown in [no active file] on line 0
The entire contents of the session will be lost. Hope this saves someone some time!
<?php // RAY_temp_ser.php
error_reporting(E_ALL);
session_start();
var_dump($_SESSION);
$_SESSION['hello'] = 'World';
var_dump($_SESSION);
// AN XML STRING FOR TEST DATA
$xml = '<?xml version="1.0"?>
<families>
<parent>
<child index="1" value="Category 1">Child One</child>
</parent>
</families>';
// MAKE AN OBJECT (GIVES SimpleXMLElement)
$obj = SimpleXML_Load_String($xml);
// STORE THE OBJECT IN THE SESSION
$_SESSION['obj'] = $obj;
__PHP_Incomplete_Class Object Demystified
1. First take note of the output. A simple example:
__PHP_Incomplete_Class Object (
[__PHP_Incomplete_Class_Name] => SomeObject1
[obj1property1] => somevalue1 [obj1property2] => __PHP_Incomplete_Class Object ( [__PHP_Incomplete_Class_Name] => SomeObject2 [obj2property1] => somevalue1 [obj2property2] => Array (
['key1'] => somevalue3, ['key2'] => somevalue4 ) ) )
2. We analyze this and break it down.
__PHP_Incomplete_Class Object tells you there is an object that needs to be declared somehow.
__PHP_Incomplete_Class_Name simply tells you the expected class name. It is just one of the properties for now.
So we have:
a) an unknown object that has a class name SomeObject1 (first class)
b) it has 2 properties, namely obj1property1 and obj2property2
c) obj2property2 is itself an object whose class name is SomeObject2 (the second class)
d) SomeObject2 has two properties, obj2property1 and obj2property2
e) obj2property2 is an array that contains two elements
3. Now that we have an idea of the structure, we shall create class definitions based from it. We will just create properties for now, methods are not required as a minimum.
<?php
class SomeObject1 {
public $obj1property1;
public $obj1property2;
}
class SomeObject2 {
public $obj2property1;
public $obj2property2;
}
?>
4. Have that accessible to your script and it will solve the __PHP_Incomplete_Class Object problem as far as the output is concerned. Now you will have:
SomeObject1 ( [obj1property1] => somevalue1 [obj1property2] => SomeObject2 ( [obj2property1] => somevalue1 [obj2property2] => Array ( ['key1'] => somevalue3, ['key2'] => somevalue4 ) ) )
As you will notice, __PHP_Incomplete_Class Object is gone and replaced by the class name. The property __PHP_Incomplete_Class_Name is also removed.
5. As for the array property obj2property2, we can directly access that and just assume that it is an array and loop through it:
<?php
// this will be SomeObject1
$data = unserialize($serialized_data);
// this will be SomeObject2
$data2 = $data->obj1property2();
foreach($data2->obj2property2 as $key => $value):
print $key.' : '. $value .'<br>';
endforeach;
?>
Outputs:
key1 : somevalue3
key2 : somevalue4
That's it. You can add more methods on the class declarations for the given properties, provided you keep your original output as basis for the data types.
When you serialize an object of a class from a particular namespace, the namespace is recorded as part of the serialization. If you decide to change this namespace's name, it can be hard to read in old serialized objects. I.e., suppose you had serialized an object of type foo\A, you change the namespace of your project to goo but otherwise leave the class definition of A unchanged. You would like to be able to unserialize the object as goo\A, instead unserialization will only create a partial object. To fix this in the case where you don't have nested objects in your class definition, you can use the following simple rename function:
/**
* Used to change the namespace of a serialized php object (assumes doesn't
* have nested subobjects)
*
* @param string $class_name new fully qualified name with namespace
* @param string $object_string serialized object
*
* @return string serialized object with new name
*/
function renameSerializedObject($class_name, $object_string)
{
/* number of digits in the length of name of the object needs to be
less than 12 digits (probably more like 4) for this to work.
*/
$name_length = intval(substr($object_string, 2, 14));
$name_space_info_length = strlen("O:".$name_length.":") +
$name_length + 2; // 2 for quotes;
$object_string = 'O:' .
strlen($class_name) . ':"'. $class_name.'"' .
substr($object_string, $name_space_info_length);
return $object_string;
}
As mentioned in the notes, unserialize returns false in the event of an error and for boolean false. Here is the first solution mentioned, without using error handling:
<?php
function isSerialized($str) {
return ($str == serialize(false) || @unserialize($str) !== false);
}
var_dump(isSerialized('s:6:"foobar";')); // bool(true)
var_dump(isSerialized('foobar')); // bool(false)
var_dump(isSerialized('b:0;')); // bool(true)
?>
When trying to serialize or unserialize recursive arrays or otherwise linked data you might find the undocumented R data type quite useful.
If you want a array like the one produced with
<?
$a = array();
$a[0] =& $a;
?>
serialized you can store it using a string simular to this one:
<?
$a = unserialize("a:1:{i:0;R:1;}");
?>
Both sources will make $a hold an array that self-references itself in index 0.
The argument for R is the index of the created sub-variable of the serialize-string beginning with 1.
When dealing with a string which contain "\r", it seems that the length is not evaluated correctly. The following solves the problem for me :
<?php
// remove the \r caracters from the $unserialized string
$unserialized = str_replace("\r","",$unserialized);
// and then unserialize()
unserialize($unserialized);
?>
Anyone having trouble serializing data with SimpleXMLElement objects stored within it, check this out:
This will traverse $data looking for any children which are instances of SimpleXMLElement, and will run ->asXML() on them, turning them into a string and making them serializable. Other data will be left alone.
<?php
function exportNestedSimpleXML($data) {
if (is_scalar($data) === false) {
foreach ($data as $k => $v) {
if ($v instanceof SimpleXMLElement) {
$v = str_replace(" ","\r",$v->asXML());
} else {
$v = exportNestedSimpleXML($v);
}
if (is_array($data)) {
$data[$k] = $v;
} else if (is_object($data)) {
$data->$k = $v;
}
}
}
return $data;
}
$data = array (
"baz" => array (
"foo" => new stdClass(),
"int" => 123,
"str" => "asdf",
"bar" => new SimpleXMLElement('<?xml version="1.0" encoding="UTF-8"?><foo>bar</foo>'),
)
);
var_dump($data);
/*array(1) {
["baz"]=>
array(4) {
["foo"]=>
object(stdClass)#3 (0) {
}
["int"]=>
int(123)
["str"]=>
string(4) "asdf"
["bar"]=>
object(SimpleXMLElement)#4 (1) {
[0]=>
string(3) "bar"
}
}
}*/
var_dump(exportNestedSimpleXML($data));
/*array(1) {
["baz"]=>
array(4) {
["foo"]=>
object(stdClass)#3 (0) {
}
["int"]=>
int(123)
["str"]=>
string(4) "asdf"
["bar"]=>
string(54) "<?xml version="1.0" encoding="UTF-8"?>
<foo>bar</foo>
"
}
}
*/
?>
Be aware that if useing serialize/unserialize in a serverfarm with both 32bit and 64bit servers you can get unexpected results.
Ex: if you serialize an integer with value of 2147483648 on a 64bit system and then unserialize it on a 32bit system you will get the value -2147483648 instead. This is because an integer on 32bit cannot be above 2147483647 so it wraps.
In reply to the earlier post about having to include object definitions *before* using unserialize. There is a workaround for this.
When an object is serialized, the first bit of the string is actually the name of the class. When an unknown object is unserialized, this is maintained as a property. So if you serialize it again, you get back the exact same string as if you'd serialized the original object. Basically, to cut to the point...
If you use
$_SESSION['my_object'] = unserialize(serialize($_SESSION['my_object']))
then you get back an object of the correct type, even if the session had originally loaded it as an object of type stdClass.
You can use the following code to use the php 7 unserialize function in php 5.3 and upwards. This adds the $option argument.
<?php
namespace
{
/**
* PHP 7 unserialize function for PHP 5.3 upwards.
* Added the $option argument (allowed_classes).
* See php unserialize manual for more detail.
**/
function php7_unserialize($str, $options = array())
{
if(version_compare(PHP_VERSION, '7.0.0', '>='))
{ return unserialize($str, $options); }
$allowed_classes = isset($options['allowed_classes']) ?
$options['allowed_classes'] : true;
if(is_array($allowed_classes) || !$allowed_classes)
{
$str = preg_replace_callback(
'/(?=^|:)(O|C):\d+:"([^"]*)":(\d+):{/',
function($matches) use ($allowed_classes)
{
if(is_array($allowed_classes) &&
in_array($matches[2], $allowed_classes))
{ return $matches[0]; }
else
{
return $matches[1].':22:"__PHP_Incomplete_Class":'.
($matches[3] + 1).
':{s:27:"__PHP_Incomplete_Class_Name";'.
serialize($matches[2]);
}
},
$str
);
}
unset($allowed_classes);
return unserialize($str);
}
} // namespace
namespace my_name_space
{
/**
* Use the new php7 unserialize in your namespace without
* renaming all unserialize(...) function calls to
* php7_unserialize(...).
**/
function unserialize($str, $options = array())
{ return php7_unserialize($str, $options); }
}
?>
Please note there is a minor difference in how unserializa() works between PHP 7 and 8.
Under PHP 8, strings that are not trimmed of whitespace will emit a warning in the likes of:
PHP Warning: unserialize(): Extra data starting at offset 721 of 722 bytes in /tmp/a.php on line 4
so something like:
$s = 's:3:"bar";'."\n"
unserialize($s); # warning emitted
unserialize(trim($s)); # no warning
I was getting unserialize() Error at offset error.
If you face similar problem then use the following procedure
$auctionDetails = preg_replace('!s:(\d+):"(.*?)";!se', "'s:'.strlen('2ドル').':\"2ドル\";'", $dataArr[$i]['auction_details'] );
$auctionDetails = unserialize($auctionDetails);
This little function will check whether the serialized string is well formed.
PHP < 6 because i'd heard changes will be made in this php-intern function,
maybe it could be edited easy for it.
<?php
function wd_check_serialization( $string, &$errmsg )
{
$str = 's';
$array = 'a';
$integer = 'i';
$any = '[^}]*?';
$count = '\d+';
$content = '"(?:\\\";|.)*?";';
$open_tag = '\{';
$close_tag = '\}';
$parameter = "($str|$array|$integer|$any):($count)" . "(?:[:]($open_tag|$content)|[;])";
$preg = "/$parameter|($close_tag)/";
if( !preg_match_all( $preg, $string, $matches ) )
{
$errmsg = 'not a serialized string';
return false;
}
$open_arrays = 0;
foreach( $matches[1] AS $key => $value )
{
if( !empty( $value ) && ( $value != $array xor $value != $str xor $value != $integer ) )
{
$errmsg = 'undefined datatype';
return false;
}
if( $value == $array )
{
$open_arrays++;
if( $matches[3][$key] != '{' )
{
$errmsg = 'open tag expected';
return false;
}
}
if( $value == '' )
{
if( $matches[4][$key] != '}' )
{
$errmsg = 'close tag expected';
return false;
}
$open_arrays--;
}
if( $value == $str )
{
$aVar = ltrim( $matches[3][$key], '"' );
$aVar = rtrim( $aVar, '";' );
if( strlen( $aVar ) != $matches[2][$key] )
{
$errmsg = 'stringlen for string not match';
return false;
}
}
if( $value == $integer )
{
if( !empty( $matches[3][$key] ) )
{
$errmsg = 'unexpected data';
return false;
}
if( !is_integer( (int)$matches[2][$key] ) )
{
$errmsg = 'integer expected';
return false;
}
}
}
if( $open_arrays != 0 )
{
$errmsg = 'wrong setted arrays';
return false;
}
return true;
}
?>
Beware: file_put_contents() is racy by design and if you use it to write serialize()d data, you may end up having error such as
unserialize(): Extra data starting at offset 1000 of 1002 bytes
For details, see https://github.com/php/php-src/issues/20108
A quick note:
If you store a serialized object in a session, you have to include the class _before_ you initialize (session_start()) the session.
If serialize() is the answer, you're almost certainly asking the wrong question.
JSON is widely available. The only thing it does not do, is the very thing that makes serialization immensely dangerous. All it takes is a crafty hacker to pass a crafted payload to a supposedly 'secured' serialize call, for a database driver to be overwritten with malicious code, for example.
Recreate the object. Normally. With actual data, and a source file, not with serialize. To do otherwise is laziness bordering on malice.
To check if a string is serialized:
$blSerialized=(@unserialize($sText)||$sText=='b:0;');
For those who are looking for an efficient solution for handling controversial "FALSE", they can use this function which in case of non-unserializable string, instead of a "FALSE", throws an Exception. Vice versa it returns the unserialized variable.
<?php
/**
* @param string $serializedString
* @param array $options
* @return mixed
* @throws Exception
*/
function UnSerialize(string $serializedString, array $options = []) {
$_unserialized = @unserialize($serializedString, $options);
if ($serializedString === serialize(false) || $_unserialized !== false){
return $_unserialized;
}
throw new Exception("Non-unserializable string");
}
?>