While reading at sla.ckers.org about some ways to get a SQL injection working if your injection point is behind a “group by” and a “limit” clause, Pragmatk came up with the PROCEDURE ANALYSE operation (available on MySQL 3/4/5) I didnt knew of yet. Although it didnt quite solve the actual problem, because it seems that you cant build some dynamic parameters for the ANALYSE function so that you could build blind SQLi vectors, it does give you information about the used database, table and column names of the query you are injecting to.
So this is another way of finding table and column names on MySQL without using the information_schema tables or load_file(). Unfortunetly you will only get the names of the columns and tables in use, but at least it will make guessing easier or maybe some columns are selected but not displayed by the webapp so that you can union select them on a different position where they do get displayed.
Here is an example: Lets assume a basic SQL query you will encounter quite often:
SELECT id, name, pass FROM users WHERE id = x
while x is our injection point. Now you can use
x = 1 PROCEDURE ANALYSE()
to get all column names, including the database and table name currently selected. You will see something like this:
Depending on the webapp you will need to use LIMIT to enumerate the result of PROCEDURE ANALYSE() line by line which contains the names in the first column of each row:
x = 1 PROCEDURE ANALYSE() #get first column name
x = 1 LIMIT 1,1 PROCEDURE ANALYSE() #get second column name
x = 1 LIMIT 2,1 PROCEDURE ANALYSE() #get third column name
With that said it is neccessary that the webapp will display the first selected column, because PROCEDURE ANALYSE will reformat the whole result with its information about the columns which is normally used to identify the best datatype for this column.
Interesting operation, I wonder if there are any other I dont know of yet which can be useful in the right circumstances.