PHP input data filtering in a nutshell

over 1 year ago sobstel php comments

Application must filter all possible forms of input data, either provided by the user or obtained from environment, database systems, web services or any other external systems.

Whole filtering must be made or repeated at server side. Client-side filtering can NEVER be trusted as it can be easily bypassed by malicious user. Do not assume that form's hidden fields, checkboxes, radio buttons or selection boxes are safe.

All input data must be filtered prior to use. It must be proved to match expected form and meet actual business rules, i.e. it must be not only of correct type, size (including length and range) and syntax, but also has proper meaning (e.g. credit card number).

Filtering strategies

• Validation - checks if given value meets certain criteria. Returns boolean value: valid or not.
• Sanitization - eliminates or translates specified characters, so value meets certain criteria. Returns safe value.

Filtering approaches

• Whitelist (accept known good) - accepts only those values which match predefined set of allowed values. All others are considered invalid.
• Blacklist (reject known bad). Denies those values which match predefined set of disallowed values. All others are considered valid.

Safer approach is always whitelist.

Input data sources

Identify untrusted input sources

• $GET, $POST
• cookies ($_COOKIE)
• environmental data $SERVER, $ENV, getenv()
• sessions ($_SESSION)
• database systems
• external systems, e.g. XML input, web services
• file uploads

$REQUEST should never be used to fetch input data (it introduces similar problems to registerglobals as it practically combines GET and POST arrays as well as cookies into a single array).

Numeric data

Casting to desired numeric type, e.g. (int)$GET['id'], (float)$GET['price']. Cast forces converting string to desired numeric type. But: Casting does not work for hexadecimal numbers, octal numbers and scientific notation. is_numeric() is lower but more flexible

Range issues: * integer - if potential value exceeds max available integer then cast to float and cut everything after decimal, on 32-bit system max integer is merely 2147483647 (PHPINTMAX) * float - stores big numbers in scientific notation

Strings

Filtering strings is usually a little complex and involves using a number of various ways and functions depending on what type of data variable should represent (URI, phone number, credit card number, username, date, time, etc).

File uploads

Turn off file uploads if you don't use them (file_uploads = off).

Limit the maximum size of an uploaded file using uploadmaxfilesize directive. Be aware that postmaxsize and memory_limit may affect this setting.

Change directory where uploads are stored (uploadtmpdir = "path/to/safe/dir") as PHP uses world-readable temporary directory by default. Any user from given shared host or simply Apache process has usually permission to read and write to the files in this directory. Note that using Suhosin cookie and session data are encrypted.

$_FILES contains data for each uploaded file. It should be handled with care.

• Check if any errors has been raised (error).
• Temporary name (tmp_name) is reliable as it's autogenerated by system.
• Filename (name) should be truncated using basename() to prevent possible overwriting of files above current directory.
• MIME type (type) should be re-checked using either getimagesize() for images or fileinfo extension for other files. Value passed by browser is unreliable as it's based only on file extension.
• Size (size) can be considered safe as it doesn't depend on user input. However, it may be compared with filsize() result to ensure that file hasn't been corrupted while waiting in temporary directory or during moving to destination directory.

Prior to filtering anything, use isuploadedfile() to check whether file was actually uploaded.

All uploaded file should be moved to desired destination directory only using moveuploadedfile(). It may be necessary to call it at the beginning to avoid openbasedir and safemode restrictions.

Serialized data

Use checksum (generated by hash_hmac()) to validate serialized data. Check checksum before unserializing. Secret key used for hashing should be random and at least 10-chars long.

Output

Output is HTML, JavaScript, JSON, databases, XML, feeds, shell commands, etc/

Ensure you escape (and encode special characters into corresponding HTML entities) all data before displaying it (prior to sending the output).

Integrity checks

Integrity checks - ensure that the data has not been tampered with/corrupted and is the same as before [OWASP05]. Integrity control types: checksum, HMAC, encryption, digital signaure depending on the security level. The preferred integrity control should be at least a HMAC using SHA-256 or preferably digitally signed or encrypted using PGP [OWASP05].

References / Further Reading

• [OWASP05] OWASP Guide 2.0
• XSS (Cross Site Scripting) Cheat Sheet (http://ha.ckers.org/xss.html)
• Alshanetsky's PHP Security Guide