zulip/zproject/backends.py

622 lines
25 KiB
Python
Raw Normal View History

import logging
from typing import Any, Dict, List, Set, Tuple, Optional, Text
2017-10-24 17:59:39 +02:00
from apiclient.sample_tools import client as googleapiclient
from django_auth_ldap.backend import LDAPBackend, _LDAPUser
import django.contrib.auth
from django.contrib.auth.backends import RemoteUserBackend
from django.conf import settings
2017-10-24 17:59:39 +02:00
from django.http import HttpResponse
from oauth2client.crypt import AppIdentityError
from social_core.backends.github import GithubOAuth2, GithubOrganizationOAuth2, \
GithubTeamOAuth2
from social_core.utils import handle_http_errors
from social_core.exceptions import AuthFailed, SocialAuthBaseException
from social_django.models import DjangoStorage
from social_django.strategy import DjangoStrategy
2017-10-24 17:59:39 +02:00
from zerver.lib.actions import do_create_user
from zerver.lib.request import JsonableError
from zerver.lib.subdomains import user_matches_subdomain, get_subdomain
2017-10-24 17:59:39 +02:00
from zerver.lib.users import check_full_name
from zerver.models import UserProfile, Realm, get_user_profile_by_id, \
get_user_profile_by_email, remote_user_to_email, email_to_username, \
get_realm, get_user
2017-10-24 17:59:39 +02:00
def pad_method_dict(method_dict):
# type: (Dict[Text, bool]) -> Dict[Text, bool]
"""Pads an authentication methods dict to contain all auth backends
supported by the software, regardless of whether they are
configured on this server"""
for key in AUTH_BACKEND_NAME_MAP:
if key not in method_dict:
method_dict[key] = False
return method_dict
def auth_enabled_helper(backends_to_check, realm):
# type: (List[Text], Optional[Realm]) -> bool
if realm is not None:
enabled_method_dict = realm.authentication_methods_dict()
pad_method_dict(enabled_method_dict)
else:
enabled_method_dict = dict((method, True) for method in Realm.AUTHENTICATION_FLAGS)
pad_method_dict(enabled_method_dict)
for supported_backend in django.contrib.auth.get_backends():
for backend_name in backends_to_check:
backend = AUTH_BACKEND_NAME_MAP[backend_name]
if enabled_method_dict[backend_name] and isinstance(supported_backend, backend):
return True
return False
def ldap_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return auth_enabled_helper(['LDAP'], realm)
def email_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return auth_enabled_helper(['Email'], realm)
def password_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return ldap_auth_enabled(realm) or email_auth_enabled(realm)
def dev_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return auth_enabled_helper(['Dev'], realm)
def google_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return auth_enabled_helper(['Google'], realm)
def github_auth_enabled(realm=None):
# type: (Optional[Realm]) -> bool
return auth_enabled_helper(['GitHub'], realm)
def any_oauth_backend_enabled(realm=None):
# type: (Optional[Realm]) -> bool
"""Used by the login page process to determine whether to show the
'OR' for login with Google"""
return auth_enabled_helper(['GitHub', 'Google'], realm)
def require_email_format_usernames(realm=None):
# type: (Optional[Realm]) -> bool
if ldap_auth_enabled(realm):
if settings.LDAP_EMAIL_ATTR or settings.LDAP_APPEND_DOMAIN:
return False
return True
def common_get_active_user(email: str, realm: Realm,
return_data: Dict[str, Any]=None) -> Optional[UserProfile]:
try:
user_profile = get_user(email, realm)
except UserProfile.DoesNotExist:
# If the user doesn't have an account in the target realm, we
# check whether they might have an account in another realm,
# and if so, provide a helpful error message via
# `invalid_subdomain`.
try:
user_profile = get_user_profile_by_email(email)
except UserProfile.DoesNotExist:
return None
if return_data is not None:
return_data['invalid_subdomain'] = True
return None
if not user_profile.is_active:
if return_data is not None:
return_data['inactive_user'] = True
return None
if user_profile.realm.deactivated:
if return_data is not None:
return_data['inactive_realm'] = True
return None
return user_profile
def common_get_active_user_by_email(email, return_data=None):
# type: (Text, Optional[Dict[str, Any]]) -> Optional[UserProfile]
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
try:
user_profile = get_user_profile_by_email(email)
except UserProfile.DoesNotExist:
return None
if not user_profile.is_active:
if return_data is not None:
return_data['inactive_user'] = True
return None
if user_profile.realm.deactivated:
if return_data is not None:
return_data['inactive_realm'] = True
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return None
return user_profile
class ZulipAuthMixin:
def get_user(self, user_profile_id):
2016-08-08 09:38:50 +02:00
# type: (int) -> Optional[UserProfile]
""" Get a UserProfile object from the user_profile_id. """
try:
return get_user_profile_by_id(user_profile_id)
except UserProfile.DoesNotExist:
return None
class SocialAuthMixin(ZulipAuthMixin):
auth_backend_name = None # type: Text
def get_email_address(self, *args, **kwargs):
# type: (*Any, **Any) -> Text
raise NotImplementedError
def get_full_name(self, *args, **kwargs):
# type: (*Any, **Any) -> Text
raise NotImplementedError
def get_authenticated_user(self, *args: Any, **kwargs: Any) -> Optional[UserProfile]:
raise NotImplementedError
@handle_http_errors
def do_auth(self, *args: Any, **kwargs: Any) -> Optional[HttpResponse]:
"""
This function is called once the authentication workflow is complete.
We override this function to:
1. Inject `return_data` and `realm_subdomain` kwargs. These will be
used by `authenticate()` functions of backends to make the
decision.
2. Call the proper authentication function to get the user in
`get_authenticated_user`.
The actual decision on authentication is done in
SocialAuthMixin._common_authenticate().
SocialAuthMixin.get_authenticated_user is expected to be overridden by
the derived class to add custom logic for authenticating the user and
returning the user.
"""
kwargs['return_data'] = {}
kwargs['realm_subdomain'] = get_subdomain(self.strategy.request) # type: ignore # `strategy` comes from Python Social Auth.
user_profile = self.get_authenticated_user(*args, **kwargs)
return self.process_do_auth(user_profile, *args, **kwargs)
def authenticate(self,
realm_subdomain: Optional[Text]='',
storage: Optional[DjangoStorage]=None,
strategy: Optional[DjangoStrategy]=None,
user: Optional[Dict[str, Any]]=None,
return_data: Optional[Dict[str, Any]]=None,
response: Optional[Dict[str, Any]]=None,
backend: Optional[GithubOAuth2]=None
) -> Optional[UserProfile]:
"""
Django decides which `authenticate` to call by inspecting the
arguments. So it's better to create `authenticate` function
with well defined arguments.
Keeping this function separate so that it can easily be
overridden.
"""
if user is None:
user = {}
assert return_data is not None
assert response is not None
return self._common_authenticate(self,
realm_subdomain=realm_subdomain,
storage=storage,
strategy=strategy,
user=user,
return_data=return_data,
response=response,
backend=backend)
def _common_authenticate(self, *args, **kwargs):
2016-08-08 09:38:50 +02:00
# type: (*Any, **Any) -> Optional[UserProfile]
return_data = kwargs.get('return_data', {})
email_address = self.get_email_address(*args, **kwargs)
if not email_address:
return_data['invalid_email'] = True
return None
try:
user_profile = get_user_profile_by_email(email_address)
except UserProfile.DoesNotExist:
return_data["valid_attestation"] = True
return None
if not user_profile.is_active:
return_data["inactive_user"] = True
return None
if user_profile.realm.deactivated:
return_data["inactive_realm"] = True
return None
if not user_matches_subdomain(kwargs.get("realm_subdomain"),
user_profile):
return_data["invalid_subdomain"] = True
return None
if not auth_enabled_helper([self.auth_backend_name], user_profile.realm):
return_data["auth_backend_disabled"] = True
return None
return user_profile
def process_do_auth(self, user_profile, *args, **kwargs):
2016-08-08 09:38:50 +02:00
# type: (UserProfile, *Any, **Any) -> Optional[HttpResponse]
# These functions need to be imported here to avoid cyclic
# dependency.
from zerver.views.auth import (login_or_register_remote_user,
redirect_to_subdomain_login_url,
redirect_and_log_into_subdomain)
return_data = kwargs.get('return_data', {})
inactive_user = return_data.get('inactive_user')
inactive_realm = return_data.get('inactive_realm')
invalid_subdomain = return_data.get('invalid_subdomain')
invalid_email = return_data.get('invalid_email')
if inactive_user or inactive_realm:
# Redirect to login page. We can't send to registration
# workflow with these errors. We will redirect to login page.
return None
if invalid_email:
# In case of invalid email, we will end up on registration page.
# This seems better than redirecting to login page.
logging.warning(
"{} got invalid email argument.".format(self.auth_backend_name)
)
strategy = self.strategy # type: ignore # This comes from Python Social Auth.
request = strategy.request
email_address = self.get_email_address(*args, **kwargs)
full_name = self.get_full_name(*args, **kwargs)
2017-05-05 19:54:36 +02:00
is_signup = strategy.session_get('is_signup') == '1'
mobile_flow_otp = strategy.session_get('mobile_flow_otp')
subdomain = strategy.session_get('subdomain')
if not subdomain:
# At least in our tests, this can be None; and it's not
# clear what the exact semantics of `session_get` are or
# what values it might return. Historically we treated
# any falsy value here as the root domain, so defensively
# continue that behavior.
subdomain = Realm.SUBDOMAIN_FOR_ROOT_DOMAIN
if (subdomain == Realm.SUBDOMAIN_FOR_ROOT_DOMAIN
or mobile_flow_otp is not None):
return login_or_register_remote_user(request, email_address,
user_profile, full_name,
2017-05-05 19:54:36 +02:00
invalid_subdomain=bool(invalid_subdomain),
mobile_flow_otp=mobile_flow_otp,
2017-05-05 19:54:36 +02:00
is_signup=is_signup)
realm = get_realm(subdomain)
if realm is None:
return redirect_to_subdomain_login_url()
2017-05-05 19:54:36 +02:00
return redirect_and_log_into_subdomain(realm, full_name, email_address,
is_signup=is_signup)
def auth_complete(self, *args, **kwargs):
# type: (*Any, **Any) -> Optional[HttpResponse]
"""
Returning `None` from this function will redirect the browser
to the login page.
"""
try:
# Call the auth_complete method of social_core.backends.oauth.BaseOAuth2
return super().auth_complete(*args, **kwargs) # type: ignore # monkey-patching
except AuthFailed:
return None
except SocialAuthBaseException as e:
logging.warning(str(e))
return None
class ZulipDummyBackend(ZulipAuthMixin):
"""
Used when we want to log you in without checking any
authentication (i.e. new user registration or when otherwise
authentication has already been checked earlier in the process).
"""
def authenticate(self, username: Optional[str]=None, realm: Optional[Realm]=None,
use_dummy_backend: bool=False,
return_data: Dict[str, Any]=None) -> Optional[UserProfile]:
if use_dummy_backend:
# These are kwargs only for readability; they should never be None
assert username is not None
assert realm is not None
return common_get_active_user(username, realm, return_data)
return None
class EmailAuthBackend(ZulipAuthMixin):
"""
Email Authentication Backend
Allows a user to sign in using an email/password pair rather than
a username/password pair.
"""
def authenticate(self, username: Optional[str]=None, password: Optional[str]=None,
realm_subdomain: Optional[str]=None,
return_data: Optional[Dict[str, Any]]=None) -> Optional[UserProfile]:
""" Authenticate a user based on email address as the user name. """
if username is None or password is None:
# Return immediately. Otherwise we will look for a SQL row with
# NULL username. While that's probably harmless, it's needless
# exposure.
return None
user_profile = common_get_active_user_by_email(username, return_data=return_data)
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
if user_profile is None:
return None
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
if not password_auth_enabled(user_profile.realm):
if return_data is not None:
return_data['password_auth_disabled'] = True
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return None
if not email_auth_enabled(user_profile.realm):
if return_data is not None:
return_data['email_auth_disabled'] = True
return None
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
if user_profile.check_password(password):
if not user_matches_subdomain(realm_subdomain, user_profile):
if return_data is not None:
return_data["invalid_subdomain"] = True
return None
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return user_profile
return None
class GoogleMobileOauth2Backend(ZulipAuthMixin):
"""
Google Apps authentication for mobile devices
Allows a user to sign in using a Google-issued OAuth2 token.
Ref:
https://developers.google.com/+/mobile/android/sign-in#server-side_access_for_your_app
https://developers.google.com/accounts/docs/CrossClientAuth#offlineAccess
"""
def authenticate(self, google_oauth2_token: str=None, realm_subdomain: str=None,
return_data: Optional[Dict[str, Any]]=None) -> Optional[UserProfile]:
if return_data is None:
return_data = {}
try:
token_payload = googleapiclient.verify_id_token(google_oauth2_token, settings.GOOGLE_CLIENT_ID)
except AppIdentityError:
return None
if token_payload["email_verified"] in (True, "true"):
try:
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
user_profile = get_user_profile_by_email(token_payload["email"])
except UserProfile.DoesNotExist:
return_data["valid_attestation"] = True
return None
if not user_profile.is_active:
return_data["inactive_user"] = True
return None
if user_profile.realm.deactivated:
return_data["inactive_realm"] = True
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return None
if not user_matches_subdomain(realm_subdomain, user_profile):
return_data["invalid_subdomain"] = True
return None
if not google_auth_enabled(realm=user_profile.realm):
return_data["google_auth_disabled"] = True
return None
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return user_profile
else:
return_data["valid_attestation"] = False
return None
class ZulipRemoteUserBackend(RemoteUserBackend):
create_unknown_user = False
def authenticate(self, remote_user: Optional[str],
realm: Optional[Realm]=None) -> Optional[UserProfile]:
assert remote_user is not None
if realm is None:
return None
if not auth_enabled_helper(["RemoteUser"], realm):
return None
email = remote_user_to_email(remote_user)
user_profile = common_get_active_user_by_email(email)
if user_profile is None:
return None
if not user_matches_subdomain(realm.subdomain, user_profile):
return None
return user_profile
class ZulipLDAPException(_LDAPUser.AuthenticationFailed):
pass
class ZulipLDAPConfigurationError(Exception):
pass
class ZulipLDAPAuthBackendBase(ZulipAuthMixin, LDAPBackend):
# Don't use Django LDAP's permissions functions
def has_perm(self, user, perm, obj=None):
# type: (Optional[UserProfile], Any, Any) -> bool
2016-08-08 09:38:50 +02:00
# Using Any type is safe because we are not doing anything with
# the arguments.
return False
def has_module_perms(self, user, app_label):
# type: (Optional[UserProfile], Optional[str]) -> bool
return False
def get_all_permissions(self, user, obj=None):
# type: (Optional[UserProfile], Any) -> Set[Any]
2016-08-08 09:38:50 +02:00
# Using Any type is safe because we are not doing anything with
# the arguments and always return empty set.
return set()
def get_group_permissions(self, user, obj=None):
# type: (Optional[UserProfile], Any) -> Set[Any]
2016-08-08 09:38:50 +02:00
# Using Any type is safe because we are not doing anything with
# the arguments and always return empty set.
return set()
def django_to_ldap_username(self, username):
# type: (Text) -> Text
if settings.LDAP_APPEND_DOMAIN:
if not username.endswith("@" + settings.LDAP_APPEND_DOMAIN):
raise ZulipLDAPException("Username does not match LDAP domain.")
return email_to_username(username)
return username
def ldap_to_django_username(self, username):
2016-08-08 09:38:50 +02:00
# type: (str) -> str
if settings.LDAP_APPEND_DOMAIN:
return "@".join((username, settings.LDAP_APPEND_DOMAIN))
return username
class ZulipLDAPAuthBackend(ZulipLDAPAuthBackendBase):
REALM_IS_NONE_ERROR = 1
def authenticate(self, username: str, password: str, realm_subdomain: Optional[Text]=None,
return_data: Optional[Dict[str, Any]]=None) -> Optional[UserProfile]:
try:
self._realm = get_realm(realm_subdomain)
username = self.django_to_ldap_username(username)
user_profile = ZulipLDAPAuthBackendBase.authenticate(self,
username=username,
password=password)
if user_profile is None:
return None
if not user_matches_subdomain(realm_subdomain, user_profile):
return None
return user_profile
except Realm.DoesNotExist:
return None # nocoverage # TODO: this may no longer be possible
except ZulipLDAPException:
return None # nocoverage # TODO: this may no longer be possible
def get_or_create_user(self, username, ldap_user):
2016-08-08 09:38:50 +02:00
# type: (str, _LDAPUser) -> Tuple[UserProfile, bool]
try:
if settings.LDAP_EMAIL_ATTR is not None:
# Get email from ldap attributes.
if settings.LDAP_EMAIL_ATTR not in ldap_user.attrs:
raise ZulipLDAPException("LDAP user doesn't have the needed %s attribute" % (
settings.LDAP_EMAIL_ATTR,))
username = ldap_user.attrs[settings.LDAP_EMAIL_ATTR][0]
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
user_profile = get_user_profile_by_email(username)
if not user_profile.is_active or user_profile.realm.deactivated:
raise ZulipLDAPException("Realm has been deactivated")
if not ldap_auth_enabled(user_profile.realm):
raise ZulipLDAPException("LDAP Authentication is not enabled")
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
return user_profile, False
except UserProfile.DoesNotExist:
if self._realm is None:
raise ZulipLDAPConfigurationError("Realm is None", self.REALM_IS_NONE_ERROR)
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
# No need to check for an inactive user since they don't exist yet
if self._realm.deactivated:
CVE-2016-4427: Fix access by deactivated realms/users. The security model for deactivated users (and users in deactivated realms) being unable to access the service is intended to work via two mechanisms: * All active user sessions are deleted, and all login code paths (where a user could get a new session) check whether the user (or realm) is inactive before authorizing the request, preventing the user from accessing the website and AJAX endpoints. * All API code paths (which don't require a session) check whether the user (and realm) are active. However, this security model was not implemented correctly. In particular, the check for whether a user has an active account in the login process was done inside the login form's validators, which meant that authentication mechanisms that did not use the login form (e.g. Google and REMOTE_USER auth) could succeed in granting a session even with an inactive account. The Zulip homepage would still fail to load because the code for / includes an API call to Tornado authorized by the user's token that would fail, but this mechanism could allow an inactive user to access realm data or users to access data in a deactivated realm. This fixes the issue by adding explicit checks for inactive users and inactive realms in all authentication backends (even those that were already protected by the login form validator). Mirror dummy users are already inactive, so we can remove the explicit code around mirror dummy users. The following commits add a complete set of tests for Zulip's inactive user and realm security model.
2016-04-21 07:19:08 +02:00
raise ZulipLDAPException("Realm has been deactivated")
full_name_attr = settings.AUTH_LDAP_USER_ATTR_MAP["full_name"]
short_name = full_name = ldap_user.attrs[full_name_attr][0]
try:
full_name = check_full_name(full_name)
except JsonableError as e:
raise ZulipLDAPException(e.msg)
if "short_name" in settings.AUTH_LDAP_USER_ATTR_MAP:
short_name_attr = settings.AUTH_LDAP_USER_ATTR_MAP["short_name"]
short_name = ldap_user.attrs[short_name_attr][0]
user_profile = do_create_user(username, None, self._realm, full_name, short_name)
return user_profile, True
# Just like ZulipLDAPAuthBackend, but doesn't let you log in.
class ZulipLDAPUserPopulator(ZulipLDAPAuthBackendBase):
def authenticate(self, username: str, password: str, realm_subdomain: Optional[Text]=None,
return_data: Optional[Dict[str, Any]]=None) -> None:
return None
class DevAuthBackend(ZulipAuthMixin):
# Allow logging in as any user without a password.
# This is used for convenience when developing Zulip.
def authenticate(self, username: str, realm_subdomain: Optional[str]=None,
return_data: Optional[Dict[str, Any]]=None) -> Optional[UserProfile]:
user_profile = common_get_active_user_by_email(username, return_data=return_data)
if user_profile is None:
return None
if not dev_auth_enabled(user_profile.realm):
return None
return user_profile
class GitHubAuthBackend(SocialAuthMixin, GithubOAuth2):
auth_backend_name = "GitHub"
def get_email_address(self, *args, **kwargs):
# type: (*Any, **Any) -> Optional[Text]
try:
return kwargs['response']['email']
except KeyError: # nocoverage # TODO: investigate
return None
def get_full_name(self, *args, **kwargs):
# type: (*Any, **Any) -> Text
2017-03-09 09:45:21 +01:00
# In case of any error return an empty string. Name is used by
# the registration page to pre-populate the name field. However,
# if it is not supplied, our registration process will make sure
# that the user enters a valid name.
try:
2017-03-09 09:45:21 +01:00
name = kwargs['response']['name']
except KeyError:
2017-03-09 09:45:21 +01:00
name = ''
if name is None:
return ''
2017-03-09 09:45:21 +01:00
return name
def get_authenticated_user(self, *args: Any, **kwargs: Any) -> Optional[UserProfile]:
"""
This function is called once the OAuth2 workflow is complete. We
override this function to call the proper `do_auth` function depending
on whether we are doing individual, team or organization based GitHub
authentication. The actual decision on authentication is done in
SocialAuthMixin._common_authenticate().
"""
user_profile = None
team_id = settings.SOCIAL_AUTH_GITHUB_TEAM_ID
org_name = settings.SOCIAL_AUTH_GITHUB_ORG_NAME
if (team_id is None and org_name is None):
try:
user_profile = GithubOAuth2.do_auth(self, *args, **kwargs)
except AuthFailed:
logging.info("User authentication failed.")
user_profile = None
elif (team_id):
backend = GithubTeamOAuth2(self.strategy, self.redirect_uri)
try:
user_profile = backend.do_auth(*args, **kwargs)
except AuthFailed:
logging.info("User is not member of GitHub team.")
user_profile = None
elif (org_name):
backend = GithubOrganizationOAuth2(self.strategy, self.redirect_uri)
try:
user_profile = backend.do_auth(*args, **kwargs)
except AuthFailed:
logging.info("User is not member of GitHub organization.")
user_profile = None
return user_profile
AUTH_BACKEND_NAME_MAP = {
'Dev': DevAuthBackend,
'Email': EmailAuthBackend,
'GitHub': GitHubAuthBackend,
'Google': GoogleMobileOauth2Backend,
'LDAP': ZulipLDAPAuthBackend,
'RemoteUser': ZulipRemoteUserBackend,
} # type: Dict[Text, Any]