PHP PDO: Protect dynamic identifiers against SQL injection - php

Currently I'm working on a concept for a simplistic tool to perform some »maintenance« database operations (drop/truncate/copy tables etc.) written in PHP.
That necessarily requires identifiers in SQL statements to be dynamic user input. While prepared statements are perfect for separating SQL statements from user input for any comparative values, they are not meant to be used for identifiers like table or column names. (That means, I can't use prepared statements to prepare identifiers.)
A common way to secure dynamic identifiers is whitelisting but this requires a static and known database structure. For example I want to implement a command like Copy table A and name it B. The interesting part here is the B.
Supposing the user is authenticated and allowed to perform this action, how can I protect this from SQL injection? Is that even possible?
I found an approach that suggests to simply quote grave accents (`) in any identifier like this:
$table_name = 'origin_table'; // can be checked against existing tables
$copy_table_name = 'user_input';
$quoted_table_name = '`' . str_replace( '`', '``', $copy_table_name ) . '`';
$sql_statement = "CREATE TABLE {$quoted_table_name} LIKE {$table_name}";
Is that a sufficient protection against possible SQL injections?
Update
PDO::quote() (mentioned in an answer) is not an option. It does not escape grave accents (`):
$user_input = 'table`; TRUNCATE TABLE users --';
var_dump( $pdo->quote( $user_input ) );
//string(33) "'table`; TRUNCATE TABLE users --'"
Update 2
The PostgreSQL extension has a function for exactly that purpose: https://secure.php.net/manual/en/function.pg-escape-identifier.php

You always want to escape ` identifiers, even table names and field names, as maybe tomorrow they will be reserved words and break your query. As long as you are using prepared statements (PDO, mysqli, or whatever) you will be fine. Note that PDO won't allow to escape table or field names.
You just need to fine grain your filtering mechanism, as to what table name, etc is allowed and not.
BTW: don't do it like you are tring to do: "CREATE TABLE {$quoted_table_name} LIKE {$table_name}";.. user prepared statements with placeholders (?, or :name, etc.)
EDIT:
Since you need to secure identifiers, I see 2 ways:
surround them with (`)
Filtering & escaping incoming values (mysql_real_escape_string or better). Filtering will include striping away any unnecessary characters (you might want to allow only letters [a-z]+ or letters with digits.

As you mention the quote() function for PDO can be used to escape both the column names as well as data. Alternatively using prepared statements such as CREATE TABLE ? LIKE ? and then just execute the query.

Related

PHP PDO protect database name

I understand that you cannot use database or table names as parameters in prepared statements. However, our app allows a user to specify the database name during the install. Is there an abstract or PDO provided way to quote these names (backticks for MySQL, brackets for MSSql, etc)?
You cannot directly escape column names and table names in PDO. You can see the answer here:
escaping column name with PDO
What you can do in this situation is to make a query to get all the tables from the given database, like this:
SHOW TABLES;
Or query to get all Databases like this one:
SHOW DATABASES;
Then use this as a white-list for the user input.
When you're using databases, it's wise to exclude some system databases like mysql itself and information_schema.
Other option is to filter the user input with a given regex, for example if your table/database names are only strings with underscore you can use:
preg_match('/^[a-z_]+$/i', $userGivenTableName)
This should remove any potential strings containing SQL Injections.

Escape a pdo query, is that necessary?

My question of to day is. Do i need to escape PDO in my script?
$columns = implode(", ",$column);
$query = ''.$query.' '.$columns.' FROM '.$table.'';
$dbh_query = $dbh->prepare($query);
$dbh_query->execute();
$dbh_querys = $dbh_query->fetchAll();
return $dbh_querys;
The whole script can be found at.
https://github.com/joshuahiwat/crud/blob/master/control/query_connector.class.php
Can someone explain why do i need a escape at this time or why not.
I like to hear from you, thanks a lot!
The parts of your query that are dynamic are the table name and column names. You can't use bind functions for these parts of the query. Bind functions can be used only for the parts of the query that would otherwise be a simple value in an SQL query. Like a numeric constant, or a quoted string or quoted date literal.
To avoid SQL injection from dynamic table names or column names, you have the following choices:
Use values that are predefined in your class, or otherwise certain to be safe. Don't use external content from users or any other source.
Use escaping. Note that the function PDO::quote() doesn't do the kind of escaping you need for table names or column names.
Create a "allowlist" of known table names and the column names for the respective table, and compare the dynamic input to the allowlist. If it doesn't match the allowlist, raise an error.
First of all you need to understand that the word you are using - "escape" - is meaningless.
What you probably mean is "to make your query safe from SQL injection". But, unfortunately, there is no such magic "escaping" that will make some abstract query safe.
The traditional query building assumes that all the query parts beside data values are hard-coded, while data values are bound via placeholders, like this:
$query = 'SELECT col1, col2 FROM some_table WHERE id = ?';
$stmt = $dbh->prepare($query);
$stmt->execute([$id]);
$row = $stmt->fetch();
This kind of a query considered safe.
In your case of a dynamically constructed query, every part is potentially vulnerable.
And here it is very important to understand that a burden of sanitizing all the query parts is entirely on this function. You cannot dismiss the danger simply claiming that your data is coming from the trusted source. That's a slippery ground because people often have no idea whether their source is trusted or not.
So, if take your question as "Do I have to protect this code from SQL injection", than the answer is - YES, YOU HAVE.
In the meantime you are protecting only a small part of your query - the data values. So you still have to protect (this term is much better than "escape") all other parts.
On a side note, your code is connecting to database every time it runs a query, which is highly inefficient and makes it impossible to use some database features.

mysql sanitize row name

I'm currently writing a php framework with focus on security. I use a query builder to generate SQL-statements, so that it is not bound to MySQL. (Or SQL in general) I found certain posibilities that user could inject row names, so it has to escape them somehow. Because of how the query builder works, i sadly cannot use prepared statements. How can I fix this?
EDIT:
The system works for example like this: db::select()-from('Tablename')->that('rowname')->run(). And I'm afraid one user could do something like that($_GET['foo']) or something. I could live with that, but I thought there has to be a way to sanatize this
To escape backtick you have to double it. Here is a function from my class
private function escapeIdent($value)
{
if ($value)
{
return "`".str_replace("`","``",$value)."`";
} else {
$this->error("Empty value for identifier (?n) placeholder");
}
}
//example:
$db->query("UPDATE users SET ?u=?s", $_POST['field'], $_POST['value']);
So, it will create a syntactically correct identifier.
But it is always better to whitelist it, as there can be a field, though with correct name,to which a user have no access rights. (So, schema-based solution is still dangerous from this point of view. Imagine there is a role field with value admin for the query from my example)
I have 2 functions in my class for this purpose, both accepts an array of allowed values.
Because of how the query builder works, i sadly cannot use prepared statements. How can I fix this?
If you can't use query parameters, then change the query builder to apply escaping to its arguments before interpolating them into SQL expressions.
Lots of people correctly advocate for query parameters, but escaping is also safe IF you do it correctly and consistently.
Cf. mysqli::real_escape_string()
Re your comment, okay I see where you're going. I was confused because you said "row name" and that's not the correct terminology. You must mean column name.
Yes, you're right, there are no functions in any of the MySQL APIs to escape table or column identifiers correctly. The escaping functions are for string literals and date literals only.
The best way to protect SQL queries when untrusted input names a table or column is to use allowlisting. That is, test the argument against a list of known table names or column names, which you either code manually, or else discover it from DESCRIBE table.
See examples of allowlisting at my past answers:
escaping column name with PDO
PHP PDO + Prepare Statement
My presentation SQL Injection Myths and Fallacies
My book SQL Antipatterns Volume 1: Avoiding the Pitfalls of Database Programming.

recommended way to safely use a user specificed order by in a SQL statement

Is it posible to bindParam in the order by portion of a sql statement.
For example, is the following possible?
select whatever from table where age > :age order by :user_specified_order_by_field_name_here
and if not, what's the recommended way to make sure that the user_specified_order_by_field_name_here does not contain SQL injection code?
No, PDO doesn't support dynamic table or column names as prepared values. Any column names you insert into the query will not be escaped, and will lead to a SQL injection vulnerability.
PDO::Quote() won't help either - it can escape strings only, but in mySQL, column names aren't strings.
The only 100% safe way to prevent problems is to compare user_specified_order_by_field_name_here against a list of valid columns in the table. You could even use numbers (that you resolve into column names internally) to add an additional layer of obscurity.

Are PHP MySQLi prepared queries with bound parameters secure?

Historically, I've always used
mysql_real_escape_string()
for all input derived from users that ends up touching the database.
Now that I've completely converted over to MySQLi and I'm using prepared queries with bound parameters, have I effectively eliminated the possibility of SQL injection attacks?
Am I correct in saying I no longer need
mysql_real_escape_string()?
This is my understanding and the basis of a project of mine:
http://sourceforge.net/projects/mysqldoneright/files/Base/MysqlDoneRight-0.23.tar.gz/download
This is not something I want to get wrong though as now that I've released it, it could affect others as well.
All user provided input will now end up in bind_parms.
The queries provided in the prepare phase are static.
Yes. Using the prepared query will escape parameters.
It's not so simple. You can use bound parameters instead of interpolating application variables into SQL expressions in place of literal values only:
$sql = "SELECT * FROM MyTable WHERE id = ".$_GET["id"]; // not safe
$sql = "SELECT * FROM MyTable WHERE id = ?"; // safe
But what if you need to make part of the query dynamic besides a literal value?
$sql = "SELECT * FROM MyTable ORDER BY ".$_GET["sortcolumn"]; // not safe
$sql = "SELECT * FROM MyTable ORDER BY ?"; // doesn't work!
The parameter will always be interpreted as a value, not a column identifier. You can run a query with ORDER BY 'score', which is different from ORDER BY score, and using a parameter will be interpreted as the former -- a constant string 'score', not the value in the column named score.
So there are lots of cases where you have to use dynamic SQL and interpolate application variables into the query to get the results you want. In those cases, query parameters can't help you. You still have to be vigilant and code defensively to prevent SQL injection flaws.
No framework or data-access library can do this work for you. You can always construct a SQL query string that contains a SQL injection flaw, and you do this before the data-access library sees the SQL query. So how is it supposed to know what's intentional and what's a flaw?
Here are the methods to achieve secure SQL queries:
Filter input. Trace any variable data that gets inserted into your SQL queries. Use input filters to strip out illegal characters. For instance, if you expect an integer, make sure the input is constrained to be an integer.
Escape output. Output in this context can be the SQL query which you send to the database server. You know you can use SQL query parameters for values, but what about a column name? You need an escaping/quoting function for identifiers, just like the old mysql_real_escape_string() is for string values.
Code reviews. Get someone to be a second pair of eyes and go over your SQL code, to help you spot places where you neglected to use the above two techniques.
When you bind parameters to a prepared statement, it escapes the data automatically, so you shouldn't escape it before you send it through. Double escaping is usually a bad thing. At the very least, it produces ugly results with extra escaped characters later on.

Categories