I'm trying to switch a web app from the "traditional" cookie-based authentication mechanism to a purely token-based one. The token should be cached once received by the client to reduce overhead. What is the best way to store the token?
This is what I already learned from googling:
The first promising avenue I looked into was browser session storage, however, as far as I can tell, this isn't shared even across tabs, which means if users follow a link from the site using a new tab, they will have to log in again.
There's also local storage, but I want users to automatically be logged out when they close their browser, and I'm also a little uneasy about tokens sitting around in the storage, even though I expire them server-side. It just seems unclean.
Another way is to store the token in a session cookie, which would mean it both gets killed on browser close and can be shared across tabs. It's almost ideal, except the cookie will of course get sent over the wire with every trip to the server, which I would like to avoid if possible. Even though it's not a security issue, it seems redundant to be sending it via the cookie as well as in the HTTP Authorization header. I've thought about setting the cookie path to a non-existent path on my domain, but that's not exactly an epitome of beauty either...
So, being faced with three non-optimal solutions, I once again turn to SO for help. How do you guys do it? What's the best way?
tl;dr What is the canonical way of persisting authentication tokens in single-page web applications?
t;dr I'm using localStorage to store token.
I'm using localStorage for storing token on client side. You can see details of my implementation in my article: React Token Based Authentication to Django REST API Backend.
The localStorage shares the token across tabs and doesn't disappear when you close the tab. The data in localStorage persist until explicitly deleted. The data in sessionStorage is deleted when the session ends.
There is some discussion on the internet that localStorage is insecure because, in the case of the XSS attack, a hacker can read all data from it. Which is true. There is some discussion that httpOnly cookies are better to store the token because hackers cant access and read the token from the httpOnly cookie. It is true, but the misunderstanding is that cookies don't prevent XSS (the XSS can occur even if you use httpOnly cookies). What is more, cookies enable CSRF attack. And, in the case of XSS, a hacker can still use set httpOnly cookies for malicious requests (similar to CSRF). So there is no clear winner here. The most secure way is to not store tokens on the client side.
Related
We decided to switch from Basic-Auth to JWT because of the session-ids that were stored in the memory and which leads sometimes to over-memory consumption in shoot-down of our Spring-Boot server that serves an Android mobile app like Twitter.
But we have some questions about JWT for our mobile use-case:
How long should an access token lives ?
How long should the refresh token lives ?
How to logout a User by invalidating his access-token and refresh token ? (For the logout we already delete the tokens on the mobile app side, but what if the tokens have being stolen by somebody and are still valid ?)
I will try to answer your queries
How long should an access token live?
You can easily configure expiry time so it depends on your requirement.
In general, try to keep it short.
How long should the refresh token live?
Above goes for refresh token with a condition that refresh token generally lives longer than access token for obvious reasons.
How to logout a User by invalidating his access-token and refresh token?
This part can be a little tricky.
You cannot manually expire a token after it has been created. So, you cannot log out with JWT on the server-side, as you do with sessions.
Can implement a few options like
When a user performs logout action or compromised. Have a blacklist which stores invalid tokens until their initial expiry date. You will need to lookup DB for every request but storage should be less as you will be storing tokens that were between logout & expiry time. You can make the blacklist efficient by keeping it in memory instead of DB.
Store Client IP Address in the claims objects JWT. When validating the token you can check with this client's IP address if it is the same source or not. You can refine it based on need like use User-Agent along with Client IP.
Worst of all reset user credentials or JWT token components to generate a new one which automatically invalidates all existing ones.
Adding a few links for more in-depth detail
Invalidate JWT Token
Destroy JWT Token
I mean it looks more like you should just be using sessions.
JWTs are not a simple replacement. They have a specific function and for some reason they have become embedded as some sort of automatic go to for any auth system.
From what you have described (the lifting of a basic auth to a more secure and modern auth system) you should be using sessions.
Good ol' Cookie sessions.
I'd go in to why more but to sum up:
A) You can control the session without odd stick on "banlist" tables and extra architecture for the JWTs for users that are banned/logged out for a system that doesn't actually need these if you just used traditional cookie based sessions.
B) They are tried and tested and the browser will keep them safe! Session cookies can be made "secure" and "http-only". There are many odd places people put JWTs including the local/session storage of a browser just waiting for a naughty js injected advert to suck them up. JWTs,just like SessionIDs, should
be in an Http-Only, Secure and Same-Site strict Cookie.
So you may as well just use a session ID and get on with life without strange front end state management when the browser is quite happy and doing that securely for you when using a Session Cookie.
C) Traditional sessions are easy to implement. Harder to understand how/why they work with all the SameSite/HttpOnly/CORS/Secure parts going on...but to implement when once understood is 99x easier and require less code when there is the Spring Framework already doing that 99% for you.
I mean sure it isn't hard to write your own JWTAuthTokenAuthFilter and implement a JWTAuthenticationProvider and a JWTCreationService and a `JWTAutoRefreshFilter...and whatever else you dream of...but why bother if you just need a session. Spring does it in like 20 lines of well tested code.
To sum up:
I mean of course properly implemented JWTs are secure...it is just maybe they are not always the best fit tool for a job.
Have a read of:
Stop Using JWTs for Sessions
Of course JWTs have a use. They are for letting a 3rd party know "yes, this is someone I know" before the client hits their API end points. Or for say having one of your servers talk to another of yours...or having client's servers talk to yours or even your servers talk to another companies:
JWT Auth - Best Practices
I have a single-page static app hosted at example.com. My server for the app is hosted at server.com. I keep these two services completely separate and they can scale differently this way. When the user wants to login their username and password is passed to yoyoma.com and a cookie is set on server.com with the access_token. The user is then redirected to example.com and is now logged in.
From the static app at example.com, we can now make AJAX requests to server.com setting withCredentials=true so that the access_token we set is passed along to server.com. This works perfectly in every browser but Safari. The only way I've gotten Safari to work is by going to preferences -> privacy -> disabling "Prevent cross-site tracking". I know that the cookies are getting set on server.com, but they don't get passed with the AJAX request. It seems to be some Privacy feature that Apple thinks is just wonderful, but how are you supposed to get around this issue. I'm not an ad service, I'm not doing anything evil, just trying to get my app to work. I specifically want to build a single page app where the server is on a different domain. Is this possible in Safari or has their privacy setting made this impossible?
Note: I should also mention to security fanatics that when the access_token cookie is set, the user is then redirected to example.com with a CSRF token. This csrf token is passed in every AJAX request by a header to prevent Cross Site Request Forgery.
Assuming you're set on keeping different domains - which is not at all unreasonable there are some ways to tackle it, but they come with compromises
1
Have example.com redirect to a server rendered server.com/login for creating the httpOnly cookie and then redirect them back to their logged in state in your SPA.
As I understand it, when Prevent cross-site tracking is enabled in Safari, the intention is to have the user interact with the other domain before cookies are allowed to be sent.
By redirecting them, that intent has been created and you should have no issue setting a cookie and have it be sent by example.com.
However it does come it its own set of limitations
Read more
2
Look into the discussion related to StorageAccess as the idea here is to work with third party auth solutions that rely on cookies from different domains. Safari ITP has sorta made those harder to use, so the idea is to work with vendors for a better solution than LocalStorage.
3
Store your key in LocalStorage, and vet all running javascript code coming from you and make sure to follow best practices when it comes to dealing with user created values.
LocalStorage exposes your to potential XSS, but Cookies expose you to CSRF attacks. You have to mitigate those and it's not too hard, but keep in mind you introduced those vectors when cookies were used.
Any XSS that happens is game over, even with httpOnly cookies. Forget your auth key, running the attackers code can do so much more damage.
Keylogger to take username/password
If the attack happens after an authed state, they could pop up a modal asking re-auth.
Submit a request to change email and then ask for password reset
Don't get me wrong, it's easier to dump LocalStorage as a generic attack versus the attacks listed above
That being said, in the end the resources you spend on dealing with cookie issues, could be better spent hardening down your javascript running on the users browsers to ensure they aren't running malicious code
You can't protect them from extensions though, and a httpOnly cookie would at the very least ensure that their keys aren't leaked. Know your compromises.
Edit: Keep in mind that whatever approach you're taking, any further reliance on browser needs to think through the different versions of browsers out in the wild. As an example, don't assume using httpOnly cookies and setting a sameSite policy negates the need for a CSRF token. Any addition of cookies might s till need a CSRF token as a header as of today unless it's a controlled environment.
Personal Opinion held loosely.
Do no harm. Cookies add "harm"/work that you have to mitigate.
Cookies might protect attacks leaking access tokens but why would an attacker care about the access token when they can run any code they wish?
LocalStorage weakness is XSS, but Cookies does not protect against that.
What prevents them from just taking the password?
Popping up a login modal to an unexpected user that won't think twice, hell Github does that from time to time for good reasons, it's not unheard of.
I have a single-page web application that uses OAuth bearer tokens to authenticate users. We do not use cookies, and there is no support for sessions. It simply makes calls to an ASP.NET Web API to access protected resources with an access token. We also support refresh tokens to obtain a new access token.
How would I implement a sliding expiration? I only see three options:
Issue a new access token on every request using the refresh token. This defeats the whole purpose of refresh tokens.
Track when the last request was in the client app. Each request would see when the last one was, and if it was after a set period, log them out and bring up the login screen. If not and their access token has expired, issue a new one and let them continue. This seems kind of messy and insecure to me.
Forget refresh tokens. Store access tokens in a database with the expiration date and update it on every request. I prefer to not do a DB operation on every request.
Is there another option or do one of these actually sound acceptable?
You said there is no session support. But this is pretty much what sessions are for, and ASP.NET and IIS support them with quite a few options for how they are managed, with or without cookies and with or without a database if I recall right. If sessions are not available in your case...
There is also the option of using an encrypted token, which contains session identity and timeout info. Then the server merely needs to know the key for decrypting the token. The server decrypts the token on each request, updates the time and sends a new encrypted token back with the new response. You can send the token as a header, cookie, part of url, take your pick. But cookies and headers are designed for this use pattern and take less work in my experience.
A token that does not decrypt is treated as an unauthorized request. Timeout is handled as you normally would, e.g. using the refresh token to get a new authentication.
If you have a server farm, only the key for decryption has to be shared between the servers. No need for a session in a database or shared cache.
You can elaborate this to expire keys over time. Then servers only have to infrequently check with a directory service, shared cache, or database, message or queue to get the most recent keys. If you generate them properly and expire them faster than someone can brute force hack them, you win! (joke) Windows has apis to support you on the encryption and key management.
I did this for a project years ago with success. It is, in effect implementing sessions without server side state. And as with all session methods and all authentication methods it has vulnerabilities.
But without some special reason to the contrary, I would just use sessions for their intended purpose. If I want each browser tab to have separate authentication I would use header based session tokens. If I want browser tabs in a browser session to share authentication I would use session cookies.
Or I would use your option three, maybe with a shared cache instead of a database, depending on performance requirements and infrastructure. I suspect that IIS+ASP.Net may even do that for you, but I have been away from them too long to know.
This might be a duplicate of this question, but the solution proposed isn't viable for us:
Protect against 3rd party callers of document.execCommand("ClearAuthenticationCache")? Clears our session cookies
Long story short: IE has a way to clear session cookies using JavaScript - document.execCommand(“ClearAuthenticationCache”). This is used in a variety of web apps including Outlook Web App (and presumably many others). Problem is MS in their infinite wisdom decided that this command should clear session cookies for all open sites (can you tell I'm a little bitter, it took me months to find the source of randomly missing JSESSIONIDs).
We use JSESSIONID as well as another token to make sure the user is authenticated. The JSESSIONID is secure and httpOnly. This works well except when the JSESSIONID is wiped out by a third party. So my question is in two parts:
Is there a way I can protect my session cookies from this (let's assume anything involving client side configuration, such as pinning or registry hacks, is a non-option)?
If not, is there a way for me to securely recover from this? Since the JSESSIONID is httpOnly, the browser shouldn't be able to read it, but maybe there is something I'm not thinking off.
If relevant: we use Tomcat 7 as our webserver. The app is a fairly complex SaaS app, and security is fairly important.
Thanks all.
I believe either of the following options would work to protect servlet sessions from document.execCommand(“ClearAuthenticationCache”):
You could set the max-age of your JSESSIONID in your web.xml. That way your JSESSIONID cookie would no longer be a session cookie! This would make your web application slightly less secure as the cookie would still survive after the browser is closed.
You could abandon HTTP cookies altogether and configure Tomcat to do session tracking with the SSL session ID. I've never actually configured it myself, but I would guess that this is more secure than using JSESSIONID cookies. However, session replication is not possible in this configuration.
I have a standard HTML login page, which I would much rather use than the standard HTTP authentication pop-up provided by browsers. Today, I am using session cookies to keep track of the session after logging in, but I'd like to be stateless and pass the HTTP authentication every time. The web services I am hitting already support this, so this is a browser-only issue.
Adding authentication credentials is trivial in jQuery, but I don't know how to keep them around. If you go from the login page (a jsp) to the Home page (another jsp) you clearly don't keep the username and password fields from the login page. I know some browsers will store your HTTP authentication credentials if you enter them from the pop-up, but I don't know if they get stored when using an XHRRequest. If they do, is there much consistency among browsers?
Also, the user needs to be able to "sign out" of the application, too. If the browser stores the authentication credentials, is there a way to clear them using JavaScript.
I feel like I can't be the first person to try to solve this. Is there some jQuery plugin or something that already handles this? Or is it simply not possible to do what I'm trying to do?
You have 2 options:
1) Client-side storage of credentials -- not a good idea. For obvious reasons you don't want to store the username/password on the client. If you had a hashed version of the password, it might not be so bad, but still not recommended. In any case, if you're going to store on the client side, you either have to use a cookie, or HTML5 local storage (which is not widely supported, yet)
2) Server-side storage of credentials -- typically done with sessions. Then the resultant Session ID can be passed back to the client and persisted in either a cookie or in the URL of each subsequent AJAX call (?SESSID=xyz for example)
The server-side approach would be the most secure, reliable, and easiest to implement
Okay, I'll take a stab at helping ...
Firstly, understand how HTTP authentication works. There are two versions - Basic and Digest. Basic transmits in plaintext, digest is encrypted. With these types of authentication, the username/password are passed in the HTTP header with every single request. The browser captures these at login and they are stored in an inaccessible browser session cookie which is deleted when the browser session is closed. So, in answer to one of your questions, you can't access these from javascript.
You could create your own session cookie variables for username and password. The jQuery functions for this are really simple. See jquery-cookie module as one example of how to set session cookies. These could be retrieved from the session cookie and sent with each ajax request and validated in the server. However, this is not a particulary good way to do authentication since sniffing the network will allow anybody to easily grab your auth details. But, it would work.
Using session cookie based authentication where the session ID is sent sent with each request is the best way to do this. At the server side, you need to have a function called for every HTTP request. This function should do the following:
check to see if the session has been authenticated
if no:
redirect to login screen
if yes:
do authorization and allow the user access to the page
Most web frameworks support session cookie authentication and the management of session ids at the server. This is definately the way to go.
This is interesting one.
Manage user sessions on server by use of cookies. Create a session when user first accesses the login page and pass the session id/key as value to one of the cookie via response. When the user is authenticated put user "key" info in cookie and "values" in application context at server. Once user is logged, any subsequent request will be authenticated based on session cookie value at server. Authorization will be done based on user "key" passed as cookie value.
On logout clear the session based cookies from server and refresh the site to default page.
Cookies are bizarre with different browsers - just a note ;)
Hope this helps.
Update
The answer below was posted in 2012 and the links are mostly dead. However, since then, a more elegant standards-based approach to the same solution appeared using JSON Web Tokens. Here is a good blog post explaining how to use them.
Most answers miss the point, which is to avoid having any server-side session. I don't want any application state in the server. I'll award the bounty to answer that came closest, but the real credit goes to the rest-discuss group and Jon Moore for the correct answer and to Mike Amundsen for helping me to actually understand it.
The best answer I've gotten is to use a cookie, but not the typical automatic session id cookie given to you by most application servers. The cookie (which will automatically be sent with each subsequent request) is a user identifier and time signed by the server. You can include an expiration time with the cookie so it simulates the typical 30 minute session on a server (which means you have to push it forward with subsequent requests) as well as keeps the same cookie from being valid forever.
The XHR/AJAX part is a red herring. This will work whether you are doing XHR requests or an old-fashioned page-by-page web application. The main points are:
The cookie is automatically sent on subsequent requests so there's no
special scripting required - it's just how browsers work already.
The server does not need to store any session for the user, so the user
can hit any server in a cluster and not have to re-authenticate.
Slightly interesting in that you consider pushing some of the authent to the client. If you want a conventional solution, KOGI's server-side suggestion is the way to go.
But you also seem to be asking questions about memory leaks involving your user supplied secrets. Good questions. But to take a general stab at answering that I'd say it would have to be browser specific. It's browser internals, javascript engine internals -dependent where a client side application (i.e., the browser, or js in the browser) is storing the values the user inputs.
Most likely those values are not replicated needlessly throughout memory, but there's no way to guarantee that. Apart from responsible javascript coding practices, there's nothing you can do to guarantee the limit of locations of user inputs.
Slight digression
The basic point is if you store it on the client it is not really secure -- unless, the serve stores encrypted information on the client with a key that only the server (or the user via their correct credentials), has. So you could conceivably code a JS application to do some authent on the client -- much the same as how bank card (used to?) do POS authent by checking the PIN to the PIN on the card, and not back at the DB. It's based on the (somewhat flimsy) assumption the user has no direct read/write access of the "dark area" cookie/local storage on client / mag strip on bank card. So I would only advise this as disqualifier for false authents and not as a sole qualifier for the credentials.
Main point
If you do want to be stateless, just store user credentials in localstorage, or as a cookie but encrypt them with a server key. When you need them send an XHR with the encrypted / use stored credentials to the server over HTTPS, let your server decrypt them and send them to the callback. Then pass those cleartext of HTTPS to do your authent.