I have two questions. I understand that if I specify the domain as .example.com (with the leading dot) in the cookie that all subdomains can share a cookie.
Can subdomain.example.com access a cookie created in example.com (without the www subdomain)?
Can example.com (without the www subdomain) access the cookie if created in subdomain.example.com?
If you set a cookie like this:
Set-Cookie: name=value
then the cookie will only apply to the request domain, and will only be sent for requests to the exact same domain, not any other subdomains. (See What is a host only cookie?)
Two different domains (e.g. example.com and subdomain.example.com, or sub1.example.com and sub2.example.com) can only share cookies if the domain attribute is present in the header:
Set-Cookie: name=value; domain=example.com
The domain attribute must "domain-match" the request URL for it to be valid, which basically means it must be the request domain or a super-domain. So this applies for both examples in the question, as well as sharing between two separate subdomains.
This cookie would then be sent for any subdomain of example.com, including nested subdomains like subsub.subdomain.example.com. (Bear in mind there are other attributes that could restrict the scope of the cookie and when it gets sent by the browser, like path or Secure).
Because of the way the domain-matching works, if you want sub1.example.com and sub2.example.com to share cookies, then you'll also share them with sub3.example.com.
See also:
www vs no-www and cookies
setcookie.net: a site where you can try it out (disclaimer: developed by me, for this question)
A note on leading dots in domain attributes: In the early RFC 2109, only domains with a leading dot (domain=.example.com) could be used across subdomains. But this could not be shared with the top-level domain, so what you ask was not possible in the older spec.
However, the newer specification RFC 6265 ignores any leading dot, meaning you can use the cookie on subdomains as well as the top-level domain.
Please everyone note that you can set a cookie from a subdomain on a domain.
(sent in the response for requesting subdomain.example.com)
Set-Cookie: name=value; Domain=example.com // GOOD
But you can't set a cookie from a domain on a subdomain.
(sent in the response for requesting example.com)
Set-Cookie: name=value; Domain=subdomain.example.com // Browser rejects cookie
Why?
According to the specifications, RFC 6265 section 5.3.6 Storage Model,
If the canonicalized request-host does not domain-match the domain-attribute: Ignore the cookie entirely and abort these steps.
and RFC 6265 section 5.1.3 Domain Matching,
Domain Matching
A string domain-matches a given domain string if at least one of the following conditions hold:
The domain string and the string are identical. (Note that both
the domain string and the string will have been canonicalized to
lower case at this point.)
All of the following conditions hold:
* The domain string is a suffix of the string.
* The last character of the string that is not included in the
domain string is a %x2E (".") character.
* The string is a host name (i.e., not an IP address).
So subdomain.example.com domain-matches example.com, but example.com does not domain-match subdomain.example.com
Check this answer also.
I'm not sure cmbuckley's answer is showing the full picture. What I read is:
Unless the cookie's attributes indicate otherwise, the cookie is
returned only to the origin server (and not, for example, to any
subdomains), and it expires at the end of the current session (as
defined by the user agent). User agents ignore unrecognized cookie.
RFC 6265
Also
8.6. Weak Integrity
Cookies do not provide integrity guarantees for sibling domains (and
their subdomains). For example, consider foo.example.com and
bar.example.com. The foo.example.com server can set a cookie with a
Domain attribute of "example.com" (possibly overwriting an existing
"example.com" cookie set by bar.example.com), and the user agent will
include that cookie in HTTP requests to bar.example.com. In the
worst case, bar.example.com will be unable to distinguish this cookie
from a cookie it set itself. The foo.example.com server might be
able to leverage this ability to mount an attack against
bar.example.com.
To me that means you can protect cookies from being read by subdomain/domain but cannot prevent writing cookies to the other domains. So somebody may rewrite your site cookies by controlling another subdomain visited by the same browser. Which might not be a big concern.
Awesome cookies test site provided by cmbuckley and for those that missed it in his answer like me; worth scrolling up.
Cookie Test
Here is an example using the DOM cookie API, so we can see the behavior for ourselves.
If we execute the following JavaScript code,
document.cookie = "key=value"
it appears to be the same as executing:
document.cookie = "key=value;domain=example.com"
The cookie key becomes available (only) on the domain example.com.
Now, if you execute the following JavaScript code on example.com,
document.cookie = "key=value;domain=.example.com"
the cookie key becomes available to example.com as well as subdomain.example.com.
Finally, if you were to try and execute the following on subdomain.example.com,
document.cookie = "key=value;domain=.example.com"
does the cookie key become available to subdomain.example.com? I was a bit surprised that this is allowed; I had assumed it would be a security violation for a subdomain to be able to set a cookie on a parent domain.
Be careful if you are working on localhost!
If you store your cookie in JavaScript like this:
document.cookie = "key=value;domain=localhost"
It might not be accessible to your subdomain, like sub.localhost. In order to solve this issue you need to use VirtualHost. For example, you can configure your virtual host with ServerName localhost.com, and then you will be able to store your cookie on your domain and subdomain like this:
document.cookie = "key=value;domain=localhost.com"
In both cases, yes, it can, and this is the default behaviour for both Internet Explorer and Edge.
The other answers add valuable insight, but they chiefly describe the behaviour in Chrome. It's important to note that the behaviour is completely different in Internet Explorer. CMBuckley's very helpful test script demonstrates that in (say) Chrome, the cookies are not shared between root and subdomains when no domain is specified.
However, the same test in Internet Explorer shows that they are shared. This Internet Explorer case is closer to the take-home description in CMBuckley's www-or-not-www link. I know this to be the case because we have a system that used different servicestack cookies on both the root and subdomain. It all worked fine until someone accessed it in Internet Explorer and the two systems fought over whose session cookie would win until we blew up the cache.
I do this and it works for me:
Cookie.set('token', 'some jwt-token', { expire:50000, domain: 'example.com' })
I have a server that stores session cookies and you can log onto it using a page (foo.com/login.html) that runs in the browser. The browser then stores a session cookie for this domain.
Now I want another page (bar.com) upon initialization to make a GET request using JavaScript to the first page (foo.com/authenticate) which should check if a session cookie exists in the browser and validate it, if correct he should respond with the session's username (however this is retrieved from the cookie). Of course I cannot check in bar.com's JavaScript if there exists a session cookie for foo.com.
Trying to solve this I ran into a few problems, one of which is of course CORS. I managed to avoid this problem by placing a reverse proxy in front of foo.com that adds all required CORS headers to the response. besides adding the headers, the proxy only tunnels requests through (eg. rev-proxy.com/authenticate -> foo.com/authenticate)
Now when I call the handler through the rev proxy from just another browser window directly (eg. rev-proxy.com/authenticate), I get the correct response. The handler from foo.com's backend finds the session cookie, reads out the username and passes it back. BUT when I try to make the same call from JavaScript inside bar.com (fetch("rev-proxy.com/authenticate")), I receive null, meaning he did not find the cookie (note that the request itself has status 200, meaning it did reach the backend of foo.com).
I have the feeling I am missing a crucial point in how cookies are used by browsers but I cannot find any useful information on my specific problem since I believe it is a rather unusual one.
See the MDN documentation:
fetch won’t send cookies, unless you set the credentials init option. (Since Aug 25, 2017. The spec changed the default credentials policy to same-origin. Firefox changed since 61.0b13.)
When setting cookie in javascript using document.cookie="dom_x=yyy;domain=www.mozilla.org;path=/", the cookie gets set in the domain www.mozilla.org.
At the same time while using document.cookie="dom_x_dot=yyy;domain:www.mozilla.org;path=/" the cookies get set in the domain .www.mozilla.org.
Is the syntax document.cookie="dom_x_dot=yyy;domain:www.mozilla.org;path=/" valid and if so when does it need to be used?
When you set a cookie domain that starts with a dot, like '.www.mozilla.org', it will be sent to that domain but also all subdomains of that domain like 'sub.www.mozilla.org'. Without the dot it will only be send to the exact domain 'www.mozilla.org'.
So I am trying to set up environment for local development to pull data from my dev server at dev.mydomain.com.
The tornado REST server serving data uses a cookie-based authentication.
To obtain the cookie I sent an AJAX post login request to the server (from the website at localhost), and the secure cookie comes back in a response. I can see that in the chrome console (network->cookies). It has the proper name, value, domain (dev.mydomain.com) and everything.
Yet, the cookie doesn't get set and the REST requests that follow fail. It is not cross-origin related. If I go to dev.mydomain.com and log in manually in another tab the cookie gets set correctly and all my subsequent requests sent from local domain work fine (since they grab the now-existent cookie).
All my requests contain this:
xhrFields: {
'withCredentials': true
}
And this is how my tornado server sets the cookie:
self.set_secure_cookie(
COOKIE_NAME, tornado.escape.url_escape(str(COOKIE_VALUE)),
expires_days=1, domain="dev.mydomain.com"
)
Any idea why the cookie doesn't get set if the login request comes from localhost?
I tried mapping 127.0.0.1 to foo.mydomain.com (for whatever that's worth) but this doesn't help.
Also, I cannot grab the cookie with javascript. Tried xhr.getResponseHeader('Set-Cookie');, yields null.
Somehow it makes sense to me that if you set the cookie for dev.mydomain.com that it does neither work for foo.mydomain.com nor for localhost.
What happens if you do something like this:
self.set_secure_cookie(
COOKIE_NAME, tornado.escape.url_escape(str(COOKIE_VALUE)),
expires_days=1, domain=".mydomain.com"
)
*.mydomain.com might work then.
EDIT:
Actually, I checked over and over again, and I can't find an example where people used the argument 'domain' for set_secure_cookie() but instead this argument exists for 'set_cookie()', as stated in the docs:
Additional keyword arguments are set on the Cookie.Morsel directly.
See http://docs.python.org/library/cookie.html#morsel-objects for
available attributes.
If you are sure about using secure cookies, you should first get sure to use a cookie secret in your application settings
class Main(web.Application):
def __init__(self):
settings = dict(
cookie_secret = "xxxx",
)
then try to set the secure cookie, without specifying the domain
self.set_secure_cookie(
COOKIE_NAME, tornado.escape.url_escape(str(COOKIE_VALUE)),
expires_days=1
)
Where can I learn (or what is) about a cookie's scope to avoid CSRF and XSS attacks for authenticated users?
For example, if I have a multi-tenant system where a single user can be access to one or more sites what is more secure:
company1.hoster.com
company2.hoster.com
company3.hoster.com
or
www.hoster.com/company1
www.hoster.com/company2
www.hoster.com/company3
What happens if I set a cookie at "hoster.com"?
You can restrict the validity scope of cookie in the domain and the path separately. So you could set a cookie in both scenarios that is only valid for that specific domain/path combination:
To set a cookie for //company1.example.com/ only:
Set-Cookie: name=value; Path=/
Omitting the Domain attribute makes the cookie only valid for the domain that it was set in. And with Path=/ the cookie is valid for any path that has the prefix /.
To set a cookie for //example.com/company1/ only:
Set-Cookie: name=value; Path=/company1/
Same explanation as for the example above. The only restriction is that you need to use /company1/ instead of /company1 as Path=/company1 would be equivalent to Path=/ and thus would make the cookie also valid for /company2 and /company3.
And to avoid that the cookie can be read via JavaScript (reducing the assets accessible using XSS), set the HttpOnly attribute.
The Open Web application security project publishes lots of valuable information about secure web application development.
Cookie's have a scope and path attributes, you would normally not want ot issue cookies for "/" or wildcard hosts *.hoster.com would both be ill-advised.
It's not as simple as this one decision, it's good you thought of security in your design, but security is a process, in every phase of your development.