HTTP
====

Scapy supports the sending / receiving of HTTP packets natively.

HTTP 1.X
--------

.. note::
    Support for HTTP 1.X was added in ``2.4.3``, whereas HTTP 2.X was already in ``2.4.0``.

About HTTP 1.X
______________

HTTP 1.X is a *text protocol*. Those are pretty unusual nowadays (HTTP 2.X is binary), therefore its implementation is very different.

For transmission purposes, HTTP 1.X frames are split in various fragments during the connection, which may or not have been encoded.
This is explain over https://developer.mozilla.org/fr/docs/Web/HTTP/Headers/Transfer-Encoding

To summarize, the frames can be split in 3 different ways:

- ``chunks``: split in fragments called chunks that are preceded by their length. The end of a frame is marked by an empty chunk
- using ``Content-Length``: the header of the HTTP frame announces the total length of the frame
- None of the above: the HTTP frame ends when the TCP stream ends / when a TCP push happens.

Moreover, each frame may be additionally compressed, depending on the algorithm specified in the HTTP header:

- ``compress``: compressed using *LZW*
- ``deflate``: compressed using *ZLIB*
- ``br``: compressed using *Brotli*
- ``gzip``

Let's have a look at what happens when you perform an HTTPRequest using Scapy's ``TCP_client`` (explained below):

.. image:: ../graphics/http_tcp.png

Once the first SYN/ACK is done, the connection is established. Scapy will send the ``HTTPRequest()``, and the host will answer with HTTP fragments. Scapy will ACK each of those, and recompile them using ``TCPSession``, like Wireshark does when it displays the answer frame.

HTTP 1.X in Scapy
_________________

Let's list the module's content::

    >>> explore(scapy.layers.http)
    Packets contained in scapy.layers.http:
    Class       |Name
    ------------|-------------
    HTTP        |HTTP 1
    HTTPRequest |HTTP Request
    HTTPResponse|HTTP Response

There are two frames available: ``HTTPRequest`` and ``HTTPResponse``. The ``HTTP`` is only used during dissection, as a util to choose between the two.
All common header fields should be supported.

- **Default HTTPRequest:**

.. code:: python

    >>> HTTPRequest().show()
    ###[ HTTP Request ]###
      Method= 'GET'
      Path= '/'
      Http_Version= 'HTTP/1.1'
      A_IM= None
      Accept= None
      Accept_Charset= None
      Accept_Datetime= None
      Accept_Encoding= None
      [...]

- **Default HTTPResponse:**

.. code:: python

    >>> HTTPResponse().show()
    ###[ HTTP Response ]###
      Http_Version= 'HTTP/1.1'
      Status_Code= '200'
      Reason_Phrase= 'OK'
      Accept_Patch43= None
      Accept_Ranges= None
      [...]

Use Scapy to send/receive HTTP 1.X
__________________________________

Scapy uses `Sessions classes <../usage.html#advanced-sniffing-sessions>`_ (more specifically the ``TCPSession`` class), in order to dissect and reconstruct HTTP packets.
This handles Content-Length, chunks and/or compression.

Here are the main ways of using HTTP 1.X with Scapy:

- :class:`~scapy.layers.http.HTTP_Client`: Automata that send HTTP requests. It supports the :func:`~scapy.layers.gssapi.SSP` mechanism to support authorization with NTLM, Kerberos, etc.
- :class:`~scapy.layers.http.HTTP_Server`: Automata to handle incoming HTTP requests. Also supports :func:`~scapy.layers.gssapi.SSP`.
- ``sniff(session=TCPSession, [...])``: Perform decompression / defragmentation on all TCP streams simultaneously, but only acts passively.
- ``TCP_client.tcplink(HTTP, host, 80)``: Acts as a raw TCP client, handles SYN/ACK, and all TCP actions, but only creates one stream. It however supports some specific features, such as changing the source IP.

**Examples:**

- :class:`~scapy.layers.http.HTTP_Client`:

Let's perform a very simple GET request to an HTTP server:

.. code:: python

    from scapy.layers.http import *  # or load_layer("http")
    client = HTTP_Client()
    resp = client.request("http://127.0.0.1:8080")
    client.close()

You can use the following shorthand to do the same very basic feature: :func:`~scapy.layers.http.http_request`, usable as so:

.. code:: python

    load_layer("http")
    http_request("www.google.com", "/")  # first argument is Host, second is Path

Let's do the same request, but this time to a server that requires NTLM authentication:

.. code:: python

    from scapy.layers.http import *  # or load_layer("http")
    client = HTTP_Client(
        HTTP_AUTH_MECHS.NTLM,
        ssp=NTLMSSP(UPN="user", PASSWORD="password"),
    )
    resp = client.request("http://127.0.0.1:8080")
    client.close()

- :class:`~scapy.layers.http.HTTP_Server`:

Start an unauthenticated HTTP server automaton:

.. code:: python

    from scapy.layers.http import *
    from scapy.layers.ntlm import *

    class Custom_HTTP_Server(HTTP_Server):
        def answer(self, pkt):
            if pkt.Path == b"/":
                return HTTPResponse() / (
                    "<!doctype html><html><body><h1>OK</h1></body></html>"
                )
            else:
                return HTTPResponse(
                    Status_Code=b"404",
                    Reason_Phrase=b"Not Found",
                ) / (
                    "<!doctype html><html><body><h1>404 - Not Found</h1></body></html>"
                )

    server = Custom_HTTP_Server.spawn(
        port=8080,
        iface="eth0",
    )

We could also have started the same server, but requiring **NTLM authorization using**:

.. code:: python

    server = HTTP_Server.spawn(
        port=8080,
        iface="eth0",
        mech=HTTP_AUTH_MECHS.NTLM,
        ssp=NTLMSSP(IDENTITIES={"user": MD4le("password")}),
    )

Or **basic auth**:

.. code:: python

    server = HTTP_Server.spawn(
        port=8080,
        iface="eth0",
        mech=HTTP_AUTH_MECHS.BASIC,
        BASIC_IDENTITIES={"user": MD4le("password")},
    )

- ``TCP_client.tcplink``:

Send an HTTPRequest to ``www.secdev.org`` and write the result in a file:

.. code:: python

    load_layer("http")
    req = HTTP()/HTTPRequest(
        Accept_Encoding=b'gzip, deflate',
        Cache_Control=b'no-cache',
        Connection=b'keep-alive',
        Host=b'www.secdev.org',
        Pragma=b'no-cache'
    )
    a = TCP_client.tcplink(HTTP, "www.secdev.org", 80)
    answer = a.sr1(req)
    a.close()
    with open("www.secdev.org.html", "wb") as file:
        file.write(answer.load)

``TCP_client.tcplink`` makes it feel like it only received one packet, but in reality it was recombined in ``TCPSession``.
If you performed a plain ``sniff()``, you would have seen those packets.

- ``sniff()``:

Dissect a pcap which contains a JPEG image that was sent over HTTP using chunks. This is able to reconstruct all HTTP streams in parallel.

.. note::

    The ``http_chunk.pcap.gz`` file is available in ``scapy/test/pcaps``

.. code:: python

    load_layer("http")
    pkts = sniff(offline="http_chunk.pcap.gz", session=TCPSession)
    # a[29] is the HTTPResponse
    with open("image.jpg", "wb") as file:
        file.write(pkts[29].load)


HTTP 2.X
--------

The HTTP 2 documentation is available as a Jupyter notebook over here: `HTTP 2 Tuto <https://github.com/secdev/scapy/blob/master/doc/notebooks/HTTP_2_Tuto.ipynb>`_