How to get response SSL certificate from requests

问题:

Trying to get the SSL certificate from a response in requests.

What is a good way to do this?

回答1:

requests deliberately wraps up low-level stuff like this. Normally, the only thing you want to do is to verify that the certs are valid. To do that, just pass verify=True. If you want to use a non-standard cacert bundle, you can pass that too. For example:

resp = requests.get('https://example.com', verify=True, cert=['/path/to/my/ca.crt'])

Also, requests is primarily a set of wrappers around other libraries, mostly urllib3 and the stdlib's http.client (or, for 2.x, httplib) and ssl.

Sometimes, the answer is just to get at the lower-level objects (e.g., resp.raw is the urllib3.response.HTTPResponse), but in many cases that's impossible.

And this is one of those cases. The only objects that ever see the certs are an http.client.HTTPSConnection (or a urllib3.connectionpool.VerifiedHTTPSConnection, but that's just a subclass of the former) and an ssl.SSLSocket, and neither of those exist anymore by the time the request returns. (As the name connectionpool implies, the HTTPSConnection object is stored in a pool, and may be reused as soon as it's done; the SSLSocket is a member of the HTTPSConnection.)

So, you need to patch things so you can copy the data up the chain. It may be as simple as this:

HTTPResponse = requests.packages.urllib3.response.HTTPResponse
orig_HTTPResponse__init__ = HTTPResponse.__init__
def new_HTTPResponse__init__(self, *args, **kwargs):
    orig_HTTPResponse__init__(self, *args, **kwargs)
    try:
        self.peercert = self._connection.sock.getpeercert()
    except AttributeError:
        pass
HTTPResponse.__init__ = new_HTTPResponse__init__

HTTPAdapter = requests.adapters.HTTPAdapter
orig_HTTPAdapter_build_response = HTTPAdapter.build_response
def new_HTTPAdapter_build_response(self, request, resp):
    response = orig_HTTPAdapter_build_response(self, request, resp)
    try:
        response.peercert = resp.peercert
    except AttributeError:
        pass
    return response
HTTPAdapter.build_response = new_HTTPAdapter_build_response

That's untested, so no guarantees; you may need to patch more than that.

Also, subclassing and overriding would probably be cleaner than monkeypatching (especially since HTTPAdapter was designed to be subclassed).

Or, even better, forking urllib3 and requests, modifying your fork, and (if you think this is legitimately useful) submitting pull requests upstream.

Anyway, now, from your code, you can do this:

resp.peercert

This will give you a dict with 'subject' and 'subjectAltName' keys, as returned by pyopenssl.WrappedSocket.getpeercert. If you instead want more information about the cert, try Christophe Vandeplas's variant of this answer that lets you get an OpenSSL.crypto.X509 object. If you want to get the entire peer certificate chain, see GoldenStake's answer.

Of course you may also want to pass along all the information necessary to verify the cert, but that's even easier, because it already passes through the top level.

回答2:

To start, abarnert's answer is very complete. While chasing the proposed connection-close issue of Kalkran I actually discovered that the peercert didn't contain detailed information about the SSL Certificate.

I dug deeper in the connection and socket info and extracted the self.sock.connection.get_peer_certificate() function which contains great functions like:

• get_subject() for CN
• get_notAfter() and get_notBefore() for expiration dates
• get_serial_number() and get_signature_algorithm() for crypto related technical details
• ...

Note that these are only available if you have pyopenssl installed on your system. Under the hood, urllib3 uses pyopenssl if it's available and the standard library's ssl module otherwise. The self.sock.connection attribute shown below only exists if self.sock is a urllib3.contrib.pyopenssl.WrappedSocket, not if it's a ssl.SSLSocket. You can install pyopenssl with pip install pyopenssl.

Once that's done, the code becomes:

import requests

HTTPResponse = requests.packages.urllib3.response.HTTPResponse
orig_HTTPResponse__init__ = HTTPResponse.__init__
def new_HTTPResponse__init__(self, *args, **kwargs):
    orig_HTTPResponse__init__(self, *args, **kwargs)
    try:
        self.peer_certificate = self._connection.peer_certificate
    except AttributeError:
        pass
HTTPResponse.__init__ = new_HTTPResponse__init__

HTTPAdapter = requests.adapters.HTTPAdapter
orig_HTTPAdapter_build_response = HTTPAdapter.build_response
def new_HTTPAdapter_build_response(self, request, resp):
    response = orig_HTTPAdapter_build_response(self, request, resp)
    try:
        response.peer_certificate = resp.peer_certificate
    except AttributeError:
        pass
    return response
HTTPAdapter.build_response = new_HTTPAdapter_build_response

HTTPSConnection = requests.packages.urllib3.connection.HTTPSConnection
orig_HTTPSConnection_connect = HTTPSConnection.connect
def new_HTTPSConnection_connect(self):
    orig_HTTPSConnection_connect(self)
    try:
        self.peer_certificate = self.sock.connection.get_peer_certificate()
    except AttributeError:
        pass
HTTPSConnection.connect = new_HTTPSConnection_connect

You will be able to access the result easily:

r = requests.get('https://yourdomain.tld', timeout=0.1)
print('Expires on: {}'.format(r.peer_certificate.get_notAfter()))
print(dir(r.peer_certificate))

If, like me, you want to ignore SSL Certificate warnings just add the following in the top of the file and do not SSL verify:

from requests.packages.urllib3.exceptions import InsecureRequestWarning
requests.packages.urllib3.disable_warnings(InsecureRequestWarning)

r = requests.get('https://yourdomain.tld', timeout=0.1, verify=False)
print(dir(r.peer_certificate))

回答3:

This, although not pretty at all, works:

import requests

req = requests.get('https://httpbin.org')
pool = req.connection.poolmanager.connection_from_url('https://httpbin.org')
conn = pool.pool.get()
# get() removes it from the pool, so put it back in
pool.pool.put(conn)
print(conn.sock.getpeercert())

回答4:

To start, abarnert's answer is very complete

But I would like to add, that in the case you're looking for the peer cert chain, you would need to patch yet another piece of code

import requests
sock_requests = requests.packages.urllib3.contrib.pyopenssl.WrappedSocket
def new_getpeercertchain(self,*args, **kwargs):
    x509 = self.connection.get_peer_cert_chain()
    return x509
sock_requests.getpeercertchain = new_getpeercertchain

after that you can call it in a very similiar manner as the accepted answer

HTTPResponse = requests.packages.urllib3.response.HTTPResponse
orig_HTTPResponse__init__ = HTTPResponse.__init__
def new_HTTPResponse__init__(self, *args, **kwargs):
    orig_HTTPResponse__init__(self, *args, **kwargs)
    try:
        self.peercertchain = self._connection.sock.getpeercertchain()
    except AttributeError:
        pass
HTTPResponse.__init__ = new_HTTPResponse__init__

HTTPAdapter = requests.adapters.HTTPAdapter
orig_HTTPAdapter_build_response = HTTPAdapter.build_response
def new_HTTPAdapter_build_response(self, request, resp):
    response = orig_HTTPAdapter_build_response(self, request, resp)
    try:
        response.peercertchain = resp.peercertchain
    except AttributeError:
        pass
    return response
HTTPAdapter.build_response = new_HTTPAdapter_build_response

you will get resp.peercertchain which contains a tuple of OpenSSL.crypto.X509 objects

回答5:

Thanks for everyone's awesome answers.

It helped me over engineer an answer to this question:

How to add a custom CA Root certificate to the CA Store used by Python in Windows?

And pull together this repository:

https://github.com/neozenith/get-ca-py

#! /usr/bin/env python
# -*- coding: utf-8 -*-
"""
Get Certificates from a request and dump them.
"""

import argparse
import sys

import requests
from requests.packages.urllib3.exceptions import InsecureRequestWarning

requests.packages.urllib3.disable_warnings(InsecureRequestWarning)

"""
Inspired by the answers from this Stackoverflow question:
https://stackoverflow.com/questions/16903528/how-to-get-response-ssl-certificate-from-requests-in-python

What follows is a series of patching the low level libraries in requests.
"""

"""
https://stackoverflow.com/a/47931103/622276
"""

sock_requests = requests.packages.urllib3.contrib.pyopenssl.WrappedSocket


def new_getpeercertchain(self, *args, **kwargs):
    x509 = self.connection.get_peer_cert_chain()
    return x509


sock_requests.getpeercertchain = new_getpeercertchain

"""
https://stackoverflow.com/a/16904808/622276
"""

HTTPResponse = requests.packages.urllib3.response.HTTPResponse
orig_HTTPResponse__init__ = HTTPResponse.__init__


def new_HTTPResponse__init__(self, *args, **kwargs):
    orig_HTTPResponse__init__(self, *args, **kwargs)
    try:
        self.peercertchain = self._connection.sock.getpeercertchain()
    except AttributeError:
        pass


HTTPResponse.__init__ = new_HTTPResponse__init__

HTTPAdapter = requests.adapters.HTTPAdapter
orig_HTTPAdapter_build_response = HTTPAdapter.build_response


def new_HTTPAdapter_build_response(self, request, resp):
    response = orig_HTTPAdapter_build_response(self, request, resp)
    try:
        response.peercertchain = resp.peercertchain
    except AttributeError:
        pass
    return response


HTTPAdapter.build_response = new_HTTPAdapter_build_response

"""
Attempt to wrap in a somewhat usable CLI
"""


def cli(args):
    parser = argparse.ArgumentParser(description="Request any URL and dump the certificate chain")
    parser.add_argument("url", metavar="URL", type=str, nargs=1, help="Valid https URL to be handled by requests")

    verify_parser = parser.add_mutually_exclusive_group(required=False)
    verify_parser.add_argument("--verify", dest="verify", action="store_true", help="Explicitly set SSL verification")
    verify_parser.add_argument(
        "--no-verify", dest="verify", action="store_false", help="Explicitly disable SSL verification"
    )
    parser.set_defaults(verify=True)

    return vars(parser.parse_args(args))


def dump_pem(cert, outfile="ca-chain.crt"):
    """Use the CN to dump certificate to PEM format"""
    PyOpenSSL = requests.packages.urllib3.contrib.pyopenssl
    pem_data = PyOpenSSL.OpenSSL.crypto.dump_certificate(PyOpenSSL.OpenSSL.crypto.FILETYPE_PEM, cert)
    issuer = cert.get_issuer().get_components()

    print(pem_data.decode("utf-8"))

    with open(outfile, "a") as output:
        for part in issuer:
            output.write(part[0].decode("utf-8"))
            output.write("=")
            output.write(part[1].decode("utf-8"))
            output.write(",\t")
        output.write("\n")
        output.write(pem_data.decode("utf-8"))


if __name__ == "__main__":
    cli_args = cli(sys.argv[1:])

    url = cli_args["url"][0]
    req = requests.get(url, verify=cli_args["verify"])
    for cert in req.peercertchain:
        dump_pem(cert)