Testing for MS Access
This article is part of the OWASP Testing Guide v3. The entire OWASP Testing Guide v3 can be downloaded here.
OWASP at the moment is working at the OWASP Testing Guide v4: you can browse the Guide here
Short Description of the Issue
In this paragraph we describe how to exploit sql injection vulnerabilties when the backend database is MS Access. In particular we focus on how to exploit blind sql injection as this kind of vulnerabilities are more frequent. After an initial introduction on which are the typical functions that are useful to exploit a sql injection vulnerability, we will introduce a clever methodology to exploit blind sql injection.
Black Box testing and example
First of all, we start showing a typical example of SQL error that we can encounter when we execute our test:
Fatal error: Uncaught exception 'com_exception' with message 'Source: Microsoft JET Database Engine
If we get this error then it's reasonable to think that we are testing an application with a MS Access Database as backend.
We have to say that, unfortunately, we start soon with a bad news since MS Access doesn't support any comment character in the sql query, so we can't use the trick of inserting the chars /* or -- or # to truncate the query. On the other hand, we can fortunately bypass this limit with the NULL character. If we insert the char %00 at some place in the query, all the remaining characters after the NULL are ignored. That happens because, internally, strings are NULL terminated. However, the NULL character can sometimes cause troubles. Luckily, 5f we try every char in the ASCII charset, we notice that there is another value that can be used in order to truncate the query. The character is 0x16 (%16 in url encoded format) or 22 in decimal. So if we have the following query:
SELECT [username],[password] FROM users WHERE [username]='$myUsername' AND [password]='$myPassword'
we can truncate the query with the following two urls:
In order to enumerate the attributes of a query, we can use the same method used for the Database MS SQL Server. In short, we can obtain the name of the attributes by error messages. For example, if we know the existence of a parameter because we got it by an error message due to the ' character, we can also know the name of the remaining attributes with the following query:
' GROUP BY Id%00
in the error message we can notice that the name of the next attribute is shown. We iterate the method until we obtain the name of all the attributes. If we don't know the name of at least one attribute, we can insert a fictitious column name, and, just like by magic, we obtain the name of the first attribute.
Obtaining Database Schema
In MS Access exist various tables that can be used to obtain the name of a table in a particular database. In the default configuration this table is not accessible, however it's worth a try. The names of these table are:
For example, if a union SQL injection vulnerability exists, you can use the following query:
' UNION SELECT Name FROM MSysObjects WHERE Type = 1%00
These are the main steps that you can use to exploit a SQL injection vulnerability on MS Access. There are also some functions that can be useful to exploit custom queries. Some of these functions are:
- ASC: Obtain the ascii value of a character passed as input
- CHR: Obtain the character of the ascii value passed as input
- LEN: Return the length of the string passed as parameter
- IIF: Is the IF construct, for example the following statement IIF(1=1, 'a', 'b') return 'a'
- MID: This function allows to extract substring, for example the following statement mid('abc',1,1) return 'a'
- TOP: This function allows to specify the maximum number of results that the query should return from the top. For example TOP 1 will return only 1 row.
- LAST: This function is used to select only the last row of a set of rows. For example the following query SELECT last(*) FROM users will return only the last row of the result.
Some of these functions will be used to exploit a blind SQL injection as we see in the next paragraph. For other functions please refer to References.
Blind sql injection testing
Blind sql injection vulnerabilities are by no mean the most frequent type of vulnerability that you will find. Generally you find a sql injection in a parameter where no union query is possible. Also, usually there is no chance to execute shell command or to read/write a file. All you can do is infer the result of your query. For our test we take the following example:
where the id parameter is used in the following query:
SELECT * FROM orders WHERE [id]=$myId
For our test we will consider the myId parameter vulnerable to blind sqli vulnerability. We want to extract the content of the table users, in particular of the column username (we have already seen how to obtain the name of the attributes thanks to the error messages and other tecniques). It is supposed that the reader alreday knowns the theory behind the blind sql injection attack, so we go straight to show some example. A typical query that can be used to infer the first character of the username of the 10th rows is:
If the first character is 'a', this query will return a 0 (a "true response"), otherwise a 'ko' string. Now we will explain why we have used this particular query. The first thing to point out is that with the functions IFF, MID and LAST we extract the first character of the username of the selected row. Unfortunately, the original query returns a set of records and not only one record, so we can't use this methodology directly. We must first select only one row. We can use the TOP function, but it only works with the first row. To select the other queries we must use a trick. We want to infer the username of the row number 10. First we use the TOP function to select the first ten rows with the query:
SELECT TOP 10 username FROM users
Then we extract from this set the last row with the function LAST. Once we have only one row and exactly the row that we want, we can use the IFF, MID and LAST function to infer the value of the username. It may be interesting to notice the use of the IFF. In our example we use IFF to return a number or a string. With this trick we can distinguish when we have a true response or not. This is because id is of numeric type, so if we compare it with a string we obtain a sql error, otherwise with the 0 value we have no errors. Of course if the parameter was of type string we can use different values. For example we can have the following query:
that returns a query that is always true if the first character is 'a' or a query that is always false in the other case (assuming that in the users table there is no id with value 'inexistenValueHere')
Thanks to this metodology we can infer the value of the username. To understand when we have obtained the complete value we have two choices:
- We try all the printable values, when no one is valid then we have the complete value.
- We can infer the length of the value (if it's a string value we can use the LEN function) and stop when we have found all the characters.
Sometimes we are blocked by some filtering function, here we see some tricks to bypass these filters.
Some filter strips away the space from the input string. We can bypass this filter using the following values as delimiter instead of the white space:
9 a c d 20 2b 2d 3d
For example we can execute the following query:
to bypass a hypothetical login form.