3 Authenticating Users
R users require local system accounts regardless of what RStudio authentication method you use. You should set up local system accounts manually and then map authenticating users to these accounts. You can also use PAM Sessions to mount your user home directory to the server.
Note: Not all RStudio products require local system accounts. Shiny Server and RStudio Connect serve end users, not R developers, so these products can be configured without local system accounts.
3.1 PAM Authentication
RStudio Server authenticates users via the Linux standard PAM (Pluggable Authentication Module) API. PAM is typically configured by default to authenticate against the system user database (
/etc/passwd) however it can also be configured to authenticate against a wide variety of other systems including ActiveDirectory and LDAP.
The section describes the PAM configuration used for authentication by default after installation. Note that PAM can be used for both authentication as well as to tailor the environment for user sessions (PAM sessions). This section describes only authentication, see the [User Resources and Limits] section for details on how RStudio Server can be configured to use PAM sessions.
3.1.1 PAM Basics
PAM profiles are are located in the
/etc/pam.d directory. Each application can have their own profile, and there is also a default profile used for applications without one (the default profile is handled differently depending on which version of Linux you are running).
To learn more about PAM and the many options and modules available for it see the following:
3.1.2 Default PAM Configuration
Debian / Ubuntu
On Debian and Ubuntu systems RStudio Server does not provide an RStudio specific PAM configuration file. As a result, RStudio Server uses the
/etc/pam.d/other profile, which by default inherits from a set of common configuration files:
@include common-auth @include common-account @include common-password @include common-session
/etc/pam.d/other profile reflects the authentication system and policies that you’d like RStudio Server to use then no further configuration is required. If you want to create a custom PAM profile for RStudio you would create a file named
/etc/pam.d/rstudio and specify whatever settings are appropriate.
RedHat / CentOS / SUSE
On RedHat, CentOS and SUSE systems applications without their own PAM profiles are denied access by default. Therefore to ensure that RStudio is running and available after installation a default PAM profile is installed at
/etc/pam.d/rstudio. This profile is configured to require a user-id greater than 500 and to authenticate users against local system accounts:
auth requisite pam_succeed_if.so uid >= 500 quiet auth required pam_unix.so nodelay account required pam_unix.so
This default PAM profile may not reflect the authentication behavior that you want for RStudio Server. In that case, some customization may be required. If you’ve already set up another PAM profile (e.g.
/etc/pam.d/login) with the desired behavior then it may be enough to simply copy that profile over the RStudio one. For example:
$ sudo cp /etc/pam.d/login /etc/pam.d/rstudio
3.1.3 Diagnosing PAM Authentication Problems
If you are unable to login to RStudio Server there may be an underlying problem with the PAM configuration. The best way to diagnose PAM configuration problems is to use the
pamtester utility (which is bundled with RStudio Server). Using
pamtester enables you to test authentication in an isolated environment as well as to see much more detailed diagnostics.
pamtester utility is located at
/usr/lib/rstudio-server/bin/pamtester. To invoke it you pass several arguments indicating the PAM profile to test, the user to test for, and whether you want verbose output. For example:
sudo /usr/lib/rstudio-server/bin/pamtester --verbose rstudio <username> authenticate
You can find more detailed documentation on using
pamtester here: http://linux.die.net/man/1/pamtester.
3.2 Restricting Access to Specific Users
3.2.1 Minimum User Id
By default RStudio Server only allows normal (as opposed to system) users to successfully authenticate. The minimum user id is determined by reading the
UID_MIN value from the
/etc/login.defs file. If the file doesn’t exist or
UID_MIN isn’t defined within it then a default value of 1000 is used.
You change the minimum user id by specifying the
auth-minimum-user-id option. For example:
Note that it’s possible that your PAM configuration is also applying a constraint on user-ids (see the Default PAM Configuration section above for an example). In this case you should ensure that the
auth-minimum-user-id is consistent with the value specified in your PAM configuration.
If your users are using very large UIDs (above 1048575/0xFFFFF), it is strongly recommended to set the
auth-minimum-user-id value to enable RStudio to make better assumptions when mapping user identifiers to projects.
3.2.2 Restricting by Group
You can specify that only users of certain groups are allowed to access RStudio Server. To do this you use the
auth-required-user-group setting. For example:
You can specify a single group as the above example does or a comma-delimited list of groups. For example:
Note that this change will not take effect until the server is restarted.
22.214.171.124 Creating and Managing Group Membership
To create a new group you use the
$ sudo groupadd <groupname>
To add a user to an existing group you use the
$ sudo usermod -a -G <groupname> <username>
Note that it’s critical that you include the
-a flag as that indicates that the group should be added to the user rather than replace the user’s group list in it’s entirety.
3.3 Google Accounts
RStudio Server can be configured to authenticate users via Google Accounts. This enables users to login with their existing Gmail or Google Apps credentials and to be automatically authenticated to RStudio Server whenever they are already logged into their Google account.
3.3.1 Registering with Google
In order to use Google Accounts with RStudio Server you need to register your server with Google for OAuth 2.0 Authentication. You do this by creating a new “Project” for your server in the Google Developer Console:
Once you’ve created a project you go to the Credentials area of APIs and auth and choose to Create New Client ID:
You’ll then be presented with a dialog used to create a new client ID:
You should select “Web application” as the application type and provide two URLs that correspond to the server you are deploying on. The screenshot above uses
https://www.example.com as the host, you should substitute your own domain and port (if not using a standard one like 80 or 443) in your configuration.
This will result in two values which you’ll need to provide as part of the RStudio Server configuration:
client-secret (they’ll be displayed in the Google Developer Console after you complete the dialog).
3.3.2 Enabling Google Accounts
To enable authentication with Google Accounts you add the
auth-google-accounts option to the RStudio Server configuration file:
In addition, you need to add a configuration file (
/etc/rstudio/google-client-secret) containing the
client-secret that you received when registering your site with Google. For example, the configuration file might look like this:
/etc/rstudio/google-client-secret file should have user read/write file permissions (i.e. 0600) to protect it’s contents from other users. You can ensure this as follows:
$ sudo chmod 0600 /etc/rstudio/google-client-secret
Note that the above
client-secret aren’t the actual values you’ll use. Rather, you should substitute the values that you obtained from Google when registering your site for OAuth authentication.
Once you enable authentication with Google Accounts that becomes the exclusive means of authentication (you can’t concurrently use both PAM and Google Account authentication).
3.3.3 Translating to Local Accounts
126.96.36.199 Creating Matching Accounts
Once a user is authenticated via Google Accounts it’s necessary to map their Google Accounts identity to a local system account. The default and most straightforward way to do this is to create a local account with a username identical to their Google email address.
If you choose to create local accounts that match Google email addresses be sure to use only lowercase characters in the account name, since Google email addresses are transformed to lower-case prior to matching them to local account names.
One problem with creating local accounts that match Google email addresses is that they often contain characters that are invalid by default within Linux usernames (e.g. @ or .). On Debian/Ubuntu systems it’s possible to force the system to create a user with these characters. Here’s an example of creating a user with a username that contains typically invalid characters:
$ sudo adduser --force-badname <username>
Note that the
--force-badname option is only available on Debian/Ubuntu systems and is not available on RedHat/CentOS or SLES systems.
If the users you are creating will only be accessing the server via RStudio, you may also want to disable their ability to login as a normal interactive user and to specify that they have no password. For example:
$ sudo adduser --force-badname --disabled-login --disabled-password <username>
188.8.131.52 Using an Account Mappings File
Alternatively, you map create local accounts that do not match Google email addresses and then specify a mapping of Google accounts to local accounts via the
/etc/rstudio/google-accounts configuration file. For example:
Note that changes to the
google-accounts configuration file take effect immediately and do not require a server restart.
3.3.4 Proxy Considerations
If you are running RStudio behind a proxy, you will need to configure your proxy to set its host name in the
X-Forwarded-Host header so that RStudio can tell the Google Web Services to redirect back to the correct location. For example, if your proxy was set up to serve RStudio requests at http://testdomain.com/rstudio/, you would want to ensure that the proxy set the
X-Forwarded-Host header to
http://testdomain.com/rstudio/. Otherwise, RStudio will attempt to redirect back to its internal address.
Alternatively, if you are running behind a proxy but cannot set the correct
X-Forwarded-Host header for whatever reason, you can use the
auth-google-accounts-redirect-base-uri option in the RStudio Server configuration file to accomplish the same purpose:
3.4 Customizing the Sign-In Page
You can customize the content and appearance of the RStudio Server sign-in page by including custom HTML within the page. This is accomplished by either:
Providing a file at
/etc/rstudio/login.htmlthat includes additional HTML to include within the login page; or
auth-login-page-htmloption within the
rserver.confconfig file which points to an alternate location for the login HTML file. For example, the following specifies that the file located at
/opt/config/rstudio-login.htmlshould be included within the login page:
3.5 Proxied Authentication
You can configure RStudio Server to participate in an existing web-based single-sign-on authentication scheme using proxied authentication. In this configuration all traffic to RStudio Server is handled by a proxy server which also handles user authentication.
In this configuration the proxy server adds a special HTTP header to requests to RStudio Server letting it know which authenticated user is making the request. RStudio Server trusts this header, launching and directing traffic to an R session owned by the specified user.
The specified user must have a local system account on the server. You should set up local system accounts manually and then map authenticating users to these accounts.
3.5.1 Enabling Proxied Authentication
To enable proxied authentication you need to specify both the
auth-proxy-sign-in-url settings (the sign-in URL is the absolute URL to the page that users should be redirected to for sign-in). For example:
Note that changes to the configuration will not take effect until the server is restarted.
3.5.2 Implementing the Proxy
184.108.40.206 Sign In URL
The sign in URL should host a page where the user specifies their credentials (this might be for example the main page for an existing web-based authentication system). After collecting and authorizing the credentials the sign in URL should then redirect back to the URL hosting the RStudio Server.
RStudio will redirect to the sign in URL under the following conditions:
Whenever an HTTP request that lacks the username header is received by the server; and
When the user clicks the “Sign out” button in the RStudio IDE user interface.
You should be sure in setting up the proxy server that traffic bound for the sign-in URL is excluded from forwarding to RStudio Server (otherwise it will end up in an infinite redirect loop).
220.127.116.11 Forwarding the Username
When proxying pre-authenticated traffic to RStudio Server you need to include a special HTTP header (by default
X-RStudio-Username) with each request indicating which user the request is associated with. For example:
It’s also possible to specify both a system username and a display username (in the case where system accounts are dynamically provisioned and don’t convey actual user identity). For example:
Note that is highly recommended that you do not use the default
X-RStudio-Username header name. The reasons for this are described in the section on security considerations immediately below.
18.104.22.168 Rewriting Usernames
It may be that the proxy system you are using sends the username in a format that doesn’t match that of users on the system, however can be easily transformed to one that does (e.g. it has a standard prefix before the username). If this is the case you can specify the
auth-proxy-user-header-rewrite option to provide a re-write rule for the inbound header. For example, the following rule strips the prefix “UID-” from a username header:
The format of a re-write rule is a regular expression followed by a space and then a replacement string. The replacement string can reference captured parts of the regular expression using
$2, etc. Consult the Boost Perl Regular Expression Syntax reference for more syntax documentation.
3.5.3 Security Considerations
22.214.171.124 Keeping the Header Name Secret
Using the the default header name
X-RStudio-Username creates a security problem: code running behind the proxy (i.e. code within R sessions) could form requests back to the server which impersonate other users (by simply inserting the header in their request).
To prevent this issue you can specify a custom header name which is kept secret from end users. This is done by creating a special configuration file (
/etc/rstudio/secure-proxy-user-header) that contains the name of the header, and then setting it’s file permissions so that it’s not readable by normal users. For example:
sudo sh -c "echo 'X-Secret-User-Header' > /etc/rstudio/secure-proxy-user-header" sudo chmod 0600 /etc/rstudio/secure-proxy-user-header
126.96.36.199 Preventing Remote Use of the Header
When implementing the proxy it’s important to remember that RStudio Server will always trust the username header to authenticate users. It’s therefore critical from the standpoint of security that all requests originating from the proxy have this header set explicitly by the proxy (as opposed to allowing the header to be specified by a remote client).
188.8.131.52 Proxy HMAC Signatures
You can force RStudio to verify an HMAC signature on requests in addition to the secret header, further ensuring the security of your system. Requiring verification of request signatures will force RStudio to only allow valid requests coming from the proxy to be authorized.
To require HMAC signatures, add the following to your
/etc/rstudio/rserver.conf configuration file:
This will require RStudio to ensure that all incoming requests contain a valid HMAC Signature in the
X-Proxy-Signature header. To properly sign requests, your proxy should compute the
HMAC SHA-256 of select fields of the request using RStudio’s secure cookie key and store the result in the
X-Proxy-Signature header. For more information on the secure cookie key, see the section on Load Balancing.
Additionally, you MUST specify the Date header on requests, which should be formatted per the HTTP standard (ex: Wed, 21 Oct 2017 07:28:00 GMT). Proper dates ALWAYS use the GMT time zone.
The string that should be signed should be constructed like the following:
X-Proxy-Signature=SHA_256(Username + "\n" + Date Header + "\n" + Request Body)
For example, assume we were proxying the following message at 11:50 AM on Tuesday, September 5th in the CDT (-5:00) time zone, and we determined that the request came from the user
GET / HTTP/1.1 User-Agent: Mozilla/4.0 (compatible; MSIE5.01; Windows NT) Host: localhost:8787 Accept-Language: en-us Accept-Encoding: gzip, deflate Connection: Keep-Alive This is a sample body!
The constructed signature would look like this:
X-Proxy-Signature=SHA_256("bdylan\nTue, 05 Sep 2017 16:50:00 GMT\nThis is a sample body!", secure_cookie_key) X-Proxy-Signature=01d90c8f2ce7de3fb75bf183fe50630cc737e2e1129ba18ea662766231e46313
Note that the above SHA_256 signature is merely an example. It will be different for every possible secure_cookie_key.
3.5.4 Troubleshooting with Access Logs
If you want to see exactly which requests RStudio Server is receiving and whether they include the expected username information, you can temporarily enable server access logs using the
server-access-log setting as follows:
After restarting RStudio Server the following file will contain a record of each HTTP request made to the server along with it’s HTTP response code:
The log file will contain entries that look like this:
127.0.0.1 - - [29/Jun/2015:06:30:41 -0400] "GET /s/f01ddf8222bea98a/ HTTP/1.1" 200 91 "http://localhost:8787/s/f01ddf8222bea98a/" "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.125 Safari/537.36" "jsmith"
Note that the very last item in the log file entry is
"jsmith". This is the username that RStudio Server read from the header passed by the proxy server. If this shows up as blank (
"-") then your proxy server isn’t forwarding the header or using the correct header name in forwarding.
Important Note: Once you’ve concluded troubleshooting it’s important that you remove the
server-access-log=1 option from the
/etc/rstudio/rserver.conf file (since this log file is not rotated it will eventually consume a large amount of disk space if you don’t remove the option).