This is a tool used to help exploit poorly designed authentication systems by locating ASCII strings that when MD5 hashed, result in raw bytes that could change SQL logic.
How It Works
When constructing SQL queries for authentication, if a prepared statement is not used – a user can perform a SQL injection attack. For example:
$query = “SELECT * FROM users WHERE email = ‘$_POST[“email”]'”
This becomes complicated though when user input is hashed, such as in the case of a password:
$email = mysql_real_escape_string($_POST[“email”]);
$pass = md5($_POST[“pass”], true);
$query = “SELECT * FROM users WHERE email = ‘$email’ AND password_hash = ‘$pass'”;
Here, the $email field is sanitized and prevents injection. But while the $pass value is not directly editable by the user, two fatal flaws exist:
- The md5() function is using the parameter raw_output = true. This results in $pass being raw bytes instead of a string containing a hex encoded representation of the hash.
- The query still is not using prepared statements for that parameter.
This results in the raw bytes of the MD5(pass) to be interpolated into the string, leaving PHP to determine encoding conversion.
HasherBasher attacks this directly. It attempts to brute force strings who’s MD5() raw result would encode to a string that would include a SQL injection to bypass authentication used by the query above.
Given the string, DyrhGOYP0vxI2DtH8y, you could calculate an MD5 hash of 6c0e97fda5c225276f522735b381a25b. But when used with raw_output = true, that looks like this:
[108 14 151 253 165 194 37 39 111 82 39 53 179 129 162 91]
In the middle of those bytes are the following:
39 111 82 39 53
‘ o R ‘ 5
So when you submit $_POST[‘pass’] with the value of DyrhGOYP0vxI2DtH8y, the query above ends up with the following logic:
SELECT * FROM users WHERE email = ‘$email’ AND password_hash = ‘…’ OR ‘5’
Which evaluates to true for the where condition, so as long as an email of a valid user, you can login as that user.
Incredibly fast. Hasherbasher is generally able to brute around 5-10 million hashes per second on standard laptops. Its speed comes from three primary sources:
- Golang’s optimized compiler and crypto library
- Parallelism via a worker pool and goroutines
- Matching has been implemented as a finite state machine
So instead of incurring the overhead of regular expressions, it’s able to locate matches orders of magnitude faster.
Effectively, the matching pattern used is:
go get github.com/gen0cide/hasherbasher
Command line options for the bruteforce subcommand are as follows:
–min-string-length value Minimum length of generated input strings
–max-string-length value Maximum length of generated input strings
–parallelism value Number of parallel brute force workers
–interval value Interval to print statistics in seconds
- interval = 5 seconds
- parallelism = number of CPUs
- min-string-length = 12
- max-string-length = 24
Author: Alex Levinson (firstname.lastname@example.org)