How to Send an HTTP Request Using HttpClient

The HttpClient mixin can be included with an exploit module in order to facilitate easier HTTP communications with a target machine.

There are mainly two common methods you will see:

send_request_raw - You use this to send a raw HTTP request. Usually, you will want this method if you need something that violates the specification; in most other cases, you should prefer send_request_cgi. If you wish to learn about how this method works, look at the documentation for Rex::Proto::Http::Client#request_raw.

Here’s a basic example of how to use send_request_raw:

	send_request_raw({'uri'=>'/index.php'})

send_request_cgi - You use this to send a more CGI-compatible HTTP request. If your request contains a query string (or POST data), then you should use this. If you wish to learn about how this method works, check out Rex::Proto::Http::Client#request_cgi.

Here’s a very basic example for send_request_cgi:

send_request_cgi({
  'method' => 'GET',
  'uri' => '/hello_world.php',
  'vars_get' => {
    'param_1' => 'abc',
    'param_2' => '123'
  }
})

Please note: send_request_raw and send_request_cgi will return a nil if there’s a timeout, so please make sure to account for that condition when you handle the return value.

Cookies & CookieJars

Part of send_request_cgi functionality is the ability to collect, edit, and send cookies via the HttpClient’s cookie_jar variable, an instance of the HttpCookieJar class.

A HttpCookieJar is a collection of HttpCookie. The Jar can be populated manually with it’s add method, or automatically via the keep_cookies option that can be passed to send_request_cgi.

If you need to clear the cookie jar (for instance, using a 2nd login), try:

cookie_jar.clear

`keep_cookies` option

Shown below is the request used to login to a gitlab account in the gitlab_file_read_rce exploit module

res = @http_client.send_request_cgi({
  'method' => 'POST',
  'uri' => '/users/sign_in',
  'keep_cookies' => true,
  'vars_post' => {
    'utf8' => '✓',
    'authenticity_token' => csrf_token,
    'user[login]' => username,
    'user[password]' => password,
    'user[remember_me]' => 0
  }
})

The cookies returned by the server with a successful login need to be attached to all future requests, so 'keep_cookies' => true, is used to add all returned cookies to the HttpClient CookieJar and attach them to all subsequent requests.

`cookie` option

Shown below is the request used to login to a gitlab account in the artica_proxy_auth_bypass_service_cmds_peform_command_injection module

artica_proxy_auth_bypass_service_cmds_peform_command_injection requires a specific cookie header to be sent with a request in order to achieve RCE. By setting a string of the desired header as the value of the cookie option, that string is set as the cookie header without any changes, allowing the exploit to be carried out.

res = send_request_cgi({
  'method' => 'GET',
  'uri' => normalize_uri(target_uri.path, 'cyrus.index.php'),
  'vars_get' => {
    'service-cmds-peform' => "||#{Rex::Text.uri_encode(cmd, 'hex-all')}||"
  },
  'cookie' => "PHPSESSID=#{@phpsessid}; AsWebStatisticsCooKie=1; shellinaboxCooKie=1"
})

Any object passed to cookie that isn’t an instance of HttpCookieJar will have to_s called on it. The result of to_s will be set as the cookie header of the http request. The contents of the HttpClient cookie_jar is ignored only this request. Subsequent requests are unaffected.

Module authors can also pass an instance of HttpCookieJar with the cookie option:

cj = Msf::Exploit::Remote::HTTP::HttpCookieJar.new

cj.add(Msf::Exploit::Remote::HTTP::HttpCookie.new('PHPSESSID', @phpsessid))
cj.add(Msf::Exploit::Remote::HTTP::HttpCookie.new('AsWebStatisticsCooKie', 1))
cj.add(Msf::Exploit::Remote::HTTP::HttpCookie.new('shellinaboxCooKie', 1))

res = send_request_cgi({
  'method' => 'GET',
  'uri' => normalize_uri(target_uri.path, 'cyrus.index.php'),
  'vars_get' => {
    'service-cmds-peform' => "||#{Rex::Text.uri_encode(cmd, 'hex-all')}||"
  },
  'cookie' => cj
})

The above code would create an identical cookie header to the one used in the previous example, save for a random ordering of the name value pairs. This shouldn’t affect how the server would read the cookies, but it’s still worth keeping in mind if you’ve somehow found a vuln reliant on the order of cookies in a header.

expire_cookies

send_request_cgi will call cleanup on cookie_jar before iot is used to populate a request with cookies. cleanup will remove any expired cookies permenetly from the jar, affecting all future requests.

If this behaviour isn’t deisred and an author would prefer to keep expired cookies in the jar, the expire_cookies option can be set to false:

res = send_request_cgi({
  'method' => 'GET',
  'uri' => normalize_uri(target_uri.path, 'cyrus.index.php'),
  'vars_get' => {
    'service-cmds-peform' => "||#{Rex::Text.uri_encode(cmd, 'hex-all')}||"
  },
  'cookie' => "PHPSESSID=#{@phpsessid}; AsWebStatisticsCooKie=1; shellinaboxCooKie=1",
  'expire_cookies' => false
})

URI Parsing

Before you send a HTTP request, you will most likely have to do some URI parsing. This is a tricky task, because sometimes when you join paths, you may accidentally get double slashes, like this: “/test//index.php”. Or for some reason you have a missing slash. These are really commonly made mistakes. So here’s how you can handle it safely:

1 - Register your default URI datastore option as ‘TARGETURI’:

Example:

register_options(
  [
    OptString.new('TARGETURI', [true, 'The base path to XXX application', '/xxx_v1/'])
  ]
)

2 - Load your TARGETURI with target_uri, that way the URI input validation will kick in, and then you get a real URI object:

In this example, we’ll just load the path:

uri = target_uri.path

3 - When you want to join another URI, always use normalize_uri:

Example:

# Returns: "/xxx_v1/admin/upload.php"
uri = normalize_uri(uri, 'admin', 'upload.php')

4 - When you’re done normalizing the URI, you’re ready to use send_request_cgi or send_request_raw

Please note: The normalize_uri method will always follow these rules:

The URI should always begin with a slash.
You will have to decide if you need the trailing slash or not.
There should be no double slashes.

Full Example

class MetasploitModule < Msf::Auxiliary

  include Msf::Exploit::Remote::HttpClient

  def initialize(info = {})
    super(
      update_info(
        info,
        'Name' => 'HttpClient Example',
        'Description' => %q{
          Do a send_request_cgi()
        },
        'Author' => [ 'sinn3r' ],
        'License' => MSF_LICENSE
      )
    )

    register_options(
      [
        OptString.new('TARGETURI', [true, 'The base path', '/'])
      ]
    )
  end

  def run
    uri = target_uri.path

    res = send_request_cgi({
      'method' => 'GET',
      'uri' => normalize_uri(uri, 'admin', 'index.php'),
      'vars_get' => {
        'p1' => 'This is param 1',
        'p2' => 'This is param 2'
      }
    })

    if res && res.code == 200
      print_good('I got a 200, awesome')
    else
      print_error('No 200, feeling blue')
    end
  end
end

Working with Burp Suite

Burp Suite is a useful tool to examine or modify HTTPS traffic while developing a module using HttpClient. To do this:

Start Burp: java -jar burpsuite.jar
In Burp, click on the Proxies tab, and then Options. Configure the proxy listener there. In this example, let’s say we have a listener on port 6666.
Once the Burp listener is up, start msfconsole and load the module you’re working on.
Enter: set Proxies HTTP:127.0.0.1:6666
Go ahead and run the module, Burp should intercept the HTTPS traffic.

Note that Burp only supports HTTPS for HttpClient. This problem is only specific to Burp and Metasploit.

If you need to examine HTTP traffic for HttpClient, a workaround is adding the following method in your module. This will override HttpClient’s send_request_* method, and return the modified output:

def send_request_cgi(opts)
  res = super(opts)
  puts res.request.to_s
  puts
  puts res.to_s
  puts
  puts
end

You can do the same for send_request_raw as well.

References

How to send an HTTP request using Rex Proto Http Client