Collectives™ on Stack Overflow

Find centralized, trusted content and collaborate around the technologies you use most.

Learn more about Collectives

Teams

Q&A for work

Connect and share knowledge within a single location that is structured and easy to search.

Learn more about Teams

I'm using requests to access a RESTful API. Everything seems to work. I can authenticate, pull back a session token and even unit test the methods in my class I wrote for the API. Then I tried to run my code.

First, here is the call I'm making. The headers are static session-related items that get set in init (). The body is build dynamically from data in a file and passed in to this function. All of the data is valid.

response = requests.post(url, headers=(Requestheader), data=json.dumps((Requestbody)))

When I run the code, it updates well over 100 records with the metadata I supply. Somewhere around item 150 I get the following:

ssl.SSLCertVerificationError: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: self signed certificate in certificate chain (_ssl.c:1045)

My first step was to call the vendor and find out if ALL of their web servers had properly signed certs figuring they were load balancing me and I found a misconfigured server. They tell me this is not the case.

Then I Googled the message and found that there is a verify kwarg, so I tried:

response = requests.post(url, headers=Requestheader, data=json.dumps(Requestbody), verify=False)

I know this isn't ideal long-term, but I wanted to test it to see if The behavior is the same. It did the same thing. It ran for a while and threw the ssl error. I thought the idea of verify=False is that it wouldn't check.

The vendor suggested I check the url I'm using, but it's fine. I would think if there were a proxy server or real man-in-the-middle attack causing problems I wouldn't see success so many times before a failure. I thought maybe it is a session timeout, but that should throw a 401 status and my activity level is too high for an inactivity timeout.

I'm a python noob and not a security professional. Suggestions appreciated.

Which version of requests? You should try to add more debug or capture the traffic to see each certificate you get, and see what changes. verify=False is in all cases not something that should be kept. – Patrick Mevzek Jul 18, 2018 at 22:35

Update

python-certifi-win32 is not maintained anymore as pointed out in the comment.

Use the drop-in library pip-system-certs instead.

If you are using Windows and you have already imported the CA in the trusted DB store, you can install the package python-certifi-win32 that automagically will use the same certificates from the Trusted DB store.

pip install python-certifi-win32

Then your code should start to work

Python Requests with wincertstore

https://gitlab.com/alelec/python-certifi-win32

Unfortunately python-certifi-win32 seems to be out-of-maintenance. Andrew Leech, the original author, suggests to move on to his other project pip-system-certs. Drop-in replacing the library with pip-system-certs (PyPI link) seems to work flawlessly. – Briareos386 Jun 16, 2022 at 17:38

I have found this over here

I found this solution, insert this code at the beginning of your source file:

import ssl
    _create_unverified_https_context = ssl._create_unverified_context
except AttributeError:
    # Legacy Python that doesn't verify HTTPS certificates by default
else:
    # Handle target environment that doesn't support HTTPS verification
    ssl._create_default_https_context = _create_unverified_https_context
                This is not advisable.  requests has a verify=False option that can turn off SSL verification.   However, that means that you are no longer checking that the machine you are talking with is the one you want to talk with.  It leaves you open to a man in the middle attack.    My final workaround was to add a retry decorator on all of my functions calling the vendor API.
– Tim B
                Mar 26, 2019 at 17:06
                Since I was inside docker talking from one container to another and didn't wanna go into adding a CA myself valid on both containers, this was exactly the solution.
– Milad.Nozari
                Apr 16, 2019 at 15:36

So the issue might have three resolutions as I see it:

  • A certificate is OK and there is something wrong with the code. The issue may occur, for example, while using prepared requests as described in this solution

    But I don't really think it is your case because in the snippet you've provided no such methods are used. For the two next variants, you'll need to get the URL that causes an error and explore it's certificate (can be done via browser).

  • A certificate is OK but a certificate authority that signed it is not included in CA list that is utilized by requests library. After you'll open a troubling URL, check CA in it and see if it's dates are valid and it is included in this list. If not, add CA in the trusted list for the requests library -- as explained in the answers to this StackOverflow question.

  • A certificate is not valid or self-singed. Same solution as in 2.

    The general solution is to wrap your script in the try except clause and to print out all the URLs that will result in mistakes. Then try to request them one by one via requests library and see if the issue occurs. If it does, it's (2) or (3) case. If not — try to run script on another machine with fresh installed python and requests. If the run will be successful — then there's some issue in your configuration.

    Thank you for the comprehensive response. I learned a few things. 1) I should be using sessions since I call a number of APIs at the site 2) It made me look again at the cert chain and it looks like there are two paths. On the second path there is a self-signed cert. ssllabs.com/ssltest/analyze.html?d=api.riscnetworks.com I'm going to look at your first link again on how to add the trust. – Tim B Jul 20, 2018 at 12:19

    For me it was sufficient just to add verify=False to the request:

    response = requests.request("POST", url, headers=headers, data=payload, verify=False)
    

    Hopefully this very easy answer helps anybody.

    This means that the SSL certificate of the HTTPS server being accessed will not be checked, and any HTTPS server (even a malicious one) could present a self-signed or invalid SSL certificate without the client being warned. – Mohammad Torkashvand Mar 13 at 9:05

    openssl s_client -showcerts -connect SERVER:443 </dev/null 2>/dev/null | sed -n -e '/BEGIN\ CERTIFICATE/,/END
    CERTIFICATE/ p' > git-mycompany-com.pem

    Install

    pip install certifi

    Execute python

    Python 3.7.3 (default, Jan 22 2021, 20:04:44) [GCC 8.3.0] on linux Type "help", "copyright", "credits" or "license" for more information.

    import certifi

    certifi.where() '/home/user/detectaEnv/lib/python3.7/site-packages/certifi/cacert.pem'

    exit()

    Execute command for add perm to python

    cat git-mycompany-com.pem | tee -a /home/user/detectaEnv/lib/python3.7/site-packages/certifi/cacert.pem

    It has worked perfectly for me

    You can set the resource URL as a trusted host by executing the below command:

    In this case i have installed the package jupyterlab

    pip install --trusted-host pypi.org --trusted-host files.pythonhosted.org jupyterlab
    
  •