How to build (and how not to build) a secure “remember me” feature | Hacker News Reader

How to build (and how not to build) a secure “remember me” feature

troyhunt.com

253 pointstroyhunt64 comments

Good article, but one thing is explained in a weird way:

Don't bother with cookie expiration. That's the wrong approach, because the cookie is controlled by the user. Always do a server-side check whether or not the auth token in the cookie is allowed to continue the session.

So you could simply set the expiration date for a cookie until 2030 but make sure that the auth token from that cookie cannot be used after $EXPIRATION_TIMESTAMP on the server.

You could also follow a layered approach:

If the user logs in every few days, re-authenticate him using the auth token from the cookie. But if the user was seen more than $MAX_INACTIVE_DAYS ago, do not re-authenticate and terminate all sessions, even if the "remember me" function is set to half a year or so.

I wouldn't go as far as "Don't bother with cookie expiration.", but you're certainly right in that you can't _rely_ on "the browser" to honor them (or even actually be a browser).

I'd still recommend setting reasonable expirations, even if it's only to be seen to be doing the right thing. Far future expirations aren't useful (as you explain) and they only serve to make it look like you're "doing in wrong". (And, for the 99.9% case of non-malicious regular users, expiring the cookies normally saves the sever the effort of looking up the session state of an already expired session with a 2030 expiry cookie. Don't _rely_ on it, but take advantage of it working right under normal conditions.)

Good point. Especially the "make it look like you're doing it wrong". You don't want to look like a data kraken in the eyes of customers.

Good point, but your explanation is also a bit confusing. If I understand correctly, your point is not so much "don't bother with cookie expiration" as "don't trust cookies to expire when you tell them to". In other words, the server should double-check cookie expiration dates because you don't want somebody fudging your 7-day cookie and using it to log in next year. Am I right?

Technically, the server cannot check cookie expiration dates, because they are never transmitted to the server. Only the cookie content is transmitted.

What I mean is: On the server, you receive an auth token that comes from the cookie. Do the logic whether or not that auth token is valid and may be used to re-authenticate the user on the server only and handle the entire logic for expiration on the server.

Do not rely on the fact that the cookie itself is present, because the user might have fiddled with the expiration date.

Implement cookie expiration server-side.

Would anyone be tempted to do this "client-side"? What would that even mean, besides what the browser does automatically? Or are we just saying, "don't trust that the client won't send expired cookies"?

> A more pragmatic mitigation is to still separate the auth cookie from a dedicated “remember me” cookie and use the latter to re-authenticate the user but impose some restrictions.

What are the benefits of using a separate cookie for the "remember me" feature, provided that you impose the same restrictions (e.g. requiring the password again before accessing sensitive areas of the website) and same security measures like "httponly" & "secure"?

I've been using a single cookie with a randomly generated and periodically replaced session identifier, which expires at the end of the session by default but lasts longer if the user selects "remember me". I'd like to know whether there is a compelling reason to switch to two cookies.

One significant benefit is being able to enforce a one-time use policy the remember-me cookie. If you only utilize it when the user is not logged in, then you use it to authenticate, set a regular session cookie, and generate a new remember-me cookie. If an old remember-me cookie is ever used that means someone probably sniffed the cookie, and you can invalidate all sessions at that point and even force a password reset if you are particularly paranoid. You can't do this with the regular session cookie because a user might have multiple tabs open. For remember-me the same effect is only a race condition if the user opens two tabs simultaneously.

> You can't do this with the regular session cookie because a user might have multiple tabs open.

Why not? It's considered standard practice to refresh regular session identifiers every X minutes, and this rarely causes race conditions unless your app is AJAX-heavy.

My apps regenerate the session identifier every time it detects that it has been more than 5 minutes since the last regeneration. So if a user who has been away for a few hours returns to the site, his session identifier will be immediately regenerated and the old one will become useless. If race conditions become a serious issue, I can allow the old session identifier to continue to work for the next 30 seconds or so. It also shouldn't be too difficult to add a feature that throws tantrums if someone continues to use the old one much later than that.

If an attacker uses a compromised session id before it is regenerated, the attacker will receive the regenerated session too. They'll have a long-lived session to the victims account.

If the attacker uses a compromised remember-me cookie, it will also be regenerated for him. Same problem.

If you use the scheme described linked from the article, when the legitimate user logs in again, the attacker will lose access to the session permanently.

Oh, I see. A separate cookie makes it easier for you to check for compromised sessions. I suppose you could also do that with regular session cookies if you keep good track of identifier history, but it'd be a lot more hassle.

The justification is earlier in the article [1].

It slightly lowers the attack cross-section. If you implement the separate login cookie scheme described in the linked article[2] you also get a limited ability to detect hijacked sessions and limit the length of time an attacker has access to the account.

[1] "One argument against long expiration of auth cookies is that they’re effectively keeping the user authenticated and at risk of attacks such as CSRF or clickjacking. Of course the application needs to exhibit other risks in order for an attacker to capitalise on the long lasting cookie but the whole defence in depth argument comes up again."

[2] http://jaspan.com/improved_persistent_login_cookie_best_prac...

I think the restrictions are things like restricting by IP, reducing simultaneous use, throttling the rate you can create new sessions.

With a unified token, you have to always trust it for as long as the 'remember me' token is alive. With split tokens, you can do a sanity checks like those mentioned above on every session initialisation.

The value of splitting the tokens into separate cookies seems to be simplicity of implementation. You can rely on the browser to invalidate the session cookie on a browser close, but leave the 'remember me' cookie (as far as you can rely on the browser to do anything).

Every useful restriction I can think of seems brittle, so I guess it's weighing up security against support.

How do you tell if the user's session has expired so you know to ask them to re-authenticate before doing anything sensitive?

AFAIK the re-authentication thing has nothing to do with expiring sessions. The whole point is to ask users to re-authenticate even though their sessions are still alive and well. Enter your password again to change your privacy settings, etc.

Remember me cookies are a specific case of some data sent to the client that the server would like to verify at a later point in time. The most direct way that comes to mind for solving this type of problem is to send the client just a random identifier and have the server look it up in a persistent store upon use.

The other way of solving it is to have the server sign whatever data is sent to the client via an HMAC. That combined with some basic serialization gets you a generic approach that you can use for all kinds of things. This also has the scalable benefit that it doesn't require a centralize persistent store.

Here's a high level summary of how we handle these use cases:

# Server -> Client (Assume server wants to round trip object X)

  1. Server serializes X to a URL friendly string (JSON + Base64 works well enough for this)
  2. Server calls generateToken(type, message, expiration) which signs the message/expiration with secretKey+type. It then returns a JSON/Base64 serialized map containing message/expiration/hmac.

# Client -> Server

  1. Client sends back the final JSON/Base64 string
  2. Server decodes and extracts the message, expiration, and hmac
  3. Server verifies the HMAC by regenerating it and comparing it against the client supplied one
  4. Server verifies the expiration date hasn't passed
  5. Server returns back the deserialized object

Couple of notes:

* The HMAC prevents the client from tampering with the message

* The 'type' field is so that tokens generated for one request type cannot be accepted somewhere else in the application. It's kind of like namespacing.

* The 'type' field does not need to be included in the message itself. It's inferred by the server based on the client request type.

* The object/message itself can be blank. In that case this becomes a secure expiring token.

Couple of possible extensions/improvements:

* If the data being sent back/forth is particularly sensitive you could also have the server encrypt either it or the entire message itself. If it's not though then it would be overkill.

* Including the requesting clients IP in the HMAC generation to further limit the set of user's it would validate against is an option as well though generally that's a bad idea. People's IP addresses change fairly often and that kills the basic use case of taking your work home with you.

This is where most cookie schemes wind up -- sending and retrieving HMACs.

I wrote my honours dissertation on an opt-in scheme that used cookies, HTTPS and javascript to track users visiting multiple websites.

What you have here is what I called "Protocol 1": the first elaboration of the naive protocol. An additional elaboration (Protocols 2 and 3) is to remove user identification from cookies entirely. Each cookie is regenerated on each HTTP request with a new ID and HMAC. This means that if a cookie is successfully harvested, its useful lifetime is limited.

There are more attacks after this. I got as far as Protocol 10; it transpires that Protocol 10 is broken anyhow. Depending on the sophistication of your attacker, there's no safe way to do what I was trying to do.

This sounds interesting. Can you provide a link to this dissertation?

Email me (in profile).

> The most direct way that comes to mind for solving this type of problem is to send the client just a random identifier and have the server look it up in a persistent store upon use.

This is my preferred approach. I generate random 128-bit integers (using a random number generator who's purpose is to be used for cryptography purposes, so it's hard to observe) and store these in Redis with a year expiry, and send them back to the client to store in a cookie. When this token in the cookie is used, the TTL on the token is reduced to 5 minutes, which lets power users up multiple tabs at once. It's the simplest, but also most robust way, I've found to deal with this.

The cleverer I try and be, the more I tend to mess these things up.

This is a good approach for a lot of things. It's the approach Flask uses for its session data by default. I wouldn't recommend it for authentication in large deployments though.

One thing that it doesn't allow for is revocation of individual tokens. Invalidating auth tokens centrally is a very useful capability to have.

Yes that's the biggest downfall and also where having a centralized persistent store wins out. The best compromise between the two I've thought of is to have a centralized store that lists revoked tokens.

Presumably there wouldn't be that many so storing the recovations alone would be more efficient. Bonus points if you add a Bloom Filter[1] in front of the revocation lookup.

In the end it's all about the use case you're solving. If the token's themselves already have a short expiration (ex: password reset token with 5 minutes to live) then revocation isn't really an issue. For something long lived and dangerous (ex: remember me token to log into my bank account) it's much more important.

[1]: http://en.wikipedia.org/wiki/Bloom_filter

Yes, a password reset token is a good example where it would be fine to used a HMACed token, and is exactly how Django handles password resets [1].

[1] https://github.com/django/django/blob/master/django/contrib/...

It's also possible to make the cookies non-exportable using a similar technique called channel binding[1], where the cookie is linked to the TLS channel it's minted over.

This is a lot more powerful than baking in the IP address when HMAC-ing the cookie but requires modification to the browser and server to get it up and running.

[1] http://www.browserauth.net/channel-bound-cookies

Regarding how long to keep the "remember me" cookie for, this is one of the overlooked reasons why native apps are eating the web's breakfast. When was the last time you have to re-authenticate on a native mobile app?

Maybe it happens on some finance/banking apps, but I don't recall seeing it on apps like Facebook or Kindle for example.

Web developers could, for example, add longer times if the user agent is mobile.

And Troy is spot on about the middle ground of "logged-in-ness". Developers are finally realising what Amazon knew all along, that it's not binary. You can keep the user partially logged in, while requiring authentication to perform sensitive tasks.

> When was the last time you have to re-authenticate on a native mobile app?

All the time. The Facebook android app is horrible.

Not for me. Did you by any chance move the app to your SD card? Apps that use the android Sync provider don't work if they're moved.

Not to refute your experience, but I've never had to re-auth facebook on my android.

Yup. Mozilla persona browser integration should already be here. Then, web apps can delegate auth to the browser much as native apps delegate to the operating system.

How does the "reference implementation" of ASP.NET actually work? Is the cookie's value mapped to some in-memory entry in the application server, is it stored in a database, etc? It seems that the article doesn't say.

> Is the cookie's value mapped to some in-memory entry in the application server, is it stored in a database, etc?

Neither. At login the server signs the new session cookie with a secret application key. Later, when a client requests a page and sends over the session cookie, the server verifies the signature. If it matches, the data in the session cookie -- which usually includes the username, whether they're logged in, and for how long -- can be trusted.

This is nice because it makes scaling a stateful site easier. Server-side session stores often become the bottleneck. Also, from a REST standpoint, updating a session database during a GET request (eg to freshen the session) is problematic.

ASP.net was the first time I remember seeing the signed session cookie trick, but most web frameworks have it now:

django -- https://docs.djangoproject.com/en/dev/topics/http/sessions/#...

rails -- http://guides.rubyonrails.org/security.html#session-storage

flask -- http://flask.pocoo.org/docs/api/#sessions

node -- https://github.com/mozilla/node-client-sessions

mojolicious -- http://toroid.org/ams/etc/mojolicious-session-cookies

Help!

Most of this discussion I understand. One point is, it all involves lots of work with browser cookies stored on the user's computer.

My approach so far is to send the user the character string version of a GUID (globally unique identifier, supposely unique in all space and time, on all computers or some such) in an HTML text box marked as "hidden" so that it doesn't show on the screen. Then I use that GUID value as a key in a key-value store where the value keeps the short term info I want on the user and their 'session'. I wrote my own little key-value store with TCP/IP, some class instance de/serialization, and two instances of a .NET collection class.

However, my understanding is that some browsers are willing to help implement the function "Remmeber me on this computer" without help of cookies. How? My guess: For the text boxes on the login in Web page, in the HTML use names for the boxes that the browser interprets as 'user name' and 'password', and then the browser stores the contents in its dedicated disk space by Web page. Right?

So, there's a 'security hole' here if the user is on a computer shared with other people!

Also, my understanding is that so far on mobile devices, the browsers do not permit cookies. True? And if so, do mobile browsers permit the function of remember me I described above?

Thanks.

You are a bit confused, but I'm not sure why you're being downvoted...

> GUID

GUIDs are really not the best idea, they can be (sort of) predicted - an attacker can guess which GUIDs are in use as session identifiers. Better to use a random string (generated by a cryptographic random number generator)

> without cookies

Browsers can store user's password when they request it. It is a security hole, consciously made by the user for added convenience. Also, only the user who made this decision can fall into this hole - it's his credentials that get stolen.

> mobile devices ... do not permit cookies

they do (although some might not store the cookies for as long as the expiration would require). However, they don't allow the user to store the password (as in the previous paragraph, the remember me function as you call it)

To round it up, it's a good thing you're experimenting with your own key/value store and session implementation, but it's usually better to use tested components in production (especially for the session stuff - it's very easy to get wrong). Anyway, keep experimenting, that's the best way to learn - I just hope somebody reviews your stuff before you shoot yourself in the foot :)

Thanks, I needed that!

I've done a lot of programming, some quite advanced, and likely for longer than 99% of the HN audience. But this is my first effort with a Web site, HTML, CSS, etc. So, I believe I can write good code and think through design issues well enough, but I am missing some information about mobile devices and browsers.

My little session state store code is working great. It was fast, fun, and easy to write, and it's been working flawlessly as I use it while developing more code.

My main alternative was to use what Microsoft's ASP.NET offered. First they offered storing session state in the same application (whatever that is, maybe just an address space). That alternative would give problems once my site is busy enough to need more than one Web server unless I had some front end box, maybe from Cisco, implement something like session affinity. The second alternative was using SQL Server. I'm trying to keep down the use of SQL Server, especially for something as light as a session state store where relational database, with ACID, etc. is gross overkill.

Thanks for the information on cookies: I don't want to assume or ask that a user's Web browser accept cookies.

When my site goes live, there will be no reason for users to have a user ID or password or log in -- later there will be. So, when I have some users log in, there is some question just how to handle that in a way both secure enough and also really easy for the users. If the user lets their Web browser store the user ID and password (not in a cookie) and a user wants to use that facility, so far fine with me -- that browser functionality is an industry standard and, thus, not to be too embarrassed about. Maybe I can get some actual documentation of that browser feature in pages for programmers at Web sites of the Web browsers -- Firefox, IE, Chrome, etc.

For using a GUID as a random key, I don't see what's wrong. I just ask for a new GUID whenever I need one. I use GUIDs for so many things inside my server code that no user can see a sequence of GUIDs from one of my servers. So, all a user could see would be just a few GUID values and should have one heck of a time guessing GUID values for other users. But, a well regarded generator of random strings would be easy to use also. I'm familiar with high quality random number generation, mostly would want assembler code for the arithmetic, don't want to try to write Intel x86 assembler to be called from Microsoft's Visual Basic .NET, would not want to use such a random number generator for session ID, and am not thrilled about writing my own random string generator. That's why for now I used GUIDs for session IDs. But I could ask ASP.NET to encrypt the GUIDs. Then a user could grab the encrypted string and maybe reuse it but would have to decrypt to see a GUID to guess the GUID sequence and extrapolate to a GUID of another connected user and then encrypt! Seems a bit much! Besides, they'd also have to guess the IP address of the other connected user and fake the IP/TCP/HTTP or whatever headers that report IP address. Maybe the NSA, some guys in a big basement in China, or some brilliant, bored wacko in Belarus wouldn't have anything better to do!

For what is stored in session state, I add to that right along as I write the rest of my Web pages. So I have a class, marked as

  <Serializable()>

and add properties to it as I want something persistent during the user's logical session.

I read somewhere that Web browsers on smart phones don't support cookies. Good to learn that this is not correct.

Since cookies are now a privacy concern, I'm still reluctant to ask my users to have cookies enabled for my site. For persistence a session ID, I don't need a cookie and can use just a GUID (or a random string) in a hidden field. For persistence for user ID and password, my site doesn't use those yet!

Thanks for the answers!

You're welcome. I had a look at your comment history and I sincerely apologize for calling you confused, I'm probably more confused than you in a lot of things. Also I appreciate your "long form" style of commenting :)

As for using either in process or SQL Server way of storing the session tokens - there is also a third way and that is writing your own session storage provider, it should be relatively easy and have all the advantages of rolling your own solution and ASP.NET session ID strength http://msdn.microsoft.com/en-us/library/ms178587(v=vs.100).a...

I don't believe you need to do anything special for the password storing (in browser) to work - just keep the same name or id on the HTML elements which accept the username and password and the URL of the login page.

An attacker can just fire as many requests as your simplest page can handle with incrementing GUIDs (not necessarily incrementing by one, GUIDs are made of some machine identification and current time so there is probably even smarter way). If any of the requests doesn't result in "Session expired" page, the attacker has just defeat your session ID because of using GUID. If you go with the custom session storage you can avoid this problem.

When you store stuff in the HttpSessionState, it gets stored (by default) in the memory of the process, it will use the custom session storage provider, if you write one.

I believe that 100% of users accept cookies. But it's true that for storing session id you don't really need it. Also in the EU we now have this stupid law requiring websites to inform users that they use cookies... http://econsultancy.com/nl/blog/9202-eu-cookie-law-three-app...

Good luck with the website - you sound like you're just the right amount of crazy for changing the world :)

"Confused" is not the worst I've been called, and at times it's, i.e., "confused", been correct!

Thanks for the feedback.

For the link to MSDN, checking my collection of such Web pages, I downloaded that page on 3/15/2012 as file number 739 from MSDN in my collection of documentation for ASP.NET, TCP/IP, etc. of the 4000+ Web pages of documentation I have so far!

Reading some of that Web page again, it looks complicated, more work to make use of their provider framework than just write my own code.

For using my code for session state store, in a Web page just write early in the execution of the Web page code

    result_code = SSS_retrieve( )

make whatever use want of

     Public users_session_state As New SPP_session

for an instance of my class

     SPP_session

and near the end of the execution of the Web page write

    result_code = SSS_store( )

Simple.

My code for making session state persistent enough, still available anywhere in my server farm with no issues about concurrency or locking, and nicely fast is really simple: So, the session state store is just a little program in Visual Basic .NET (apparently as good as anything else at making use of the common language runtime, CLR, and .NET Framework).

The program runs as just a console application started just from a command line, actually from a short scripting program (in ObjectRexx, my favorite scripting language until I convert over to Power Shell).

The program implements just a key-value store. To this program, each key is just an array of type byte; similarly for each value. The implementation is just via an instance of a .NET collection class, hopefully AVL trees or red-black trees. There is a second instance of that collection class used for handling session time outs, as I recall, hard coded at 40 minutes. We're taking simple, dirt simple, here!

The program just listens on a dedicated (hard coded!) IP port number for a connection request. Since the program is single threaded, there are no problems with concurrency or locking. That is, the program serves just one customer at a time!

From timings, on a single core, 3.0 GHz processor, the program should be able to do the work needed for a site to send 1000+ Web pages a second. If my Web site needs more capacity, then just do sharding, that is, in the code for a Web page, given a session ID, do a look up and find the IP address and port number of the executing instance of the session state store program that handles that session ID.

Keep 512 such instances busy, and could have a really busy Web site! With 16 threads and instances per processor, that would be 32 processors which would fit easily in just two standard racks. A big Web site and session state handled with just two racks! Nice!

Microsoft has a collection class in storage with scope the 'application', i.e., all Web pages in one instance of IIS, and I tried to use that storage for my session state store. The first bad news was that during development something got confused raising an exception "invalid cast". So each time I changed the code for a Web page I had to issue Microsoft command

     iisreset /restart

to get everything working again.

It is true that my way of handling session state makes my Web page code look different.

So far I'm happy with how I'm handling session state.

Thanks for the detail on how the next GUID is generated -- just add 1, which is less obscure than I had hoped.

So, maybe for a GUID start with the MAC address of a network adapter, use some time and/or date, and then increment or some such.

If a malicious user wants to break into my site, then they will have to guess both the GUID of the session ID of a current user and also that user's IP address. If that is not secure enough, then I can just ask ASP.NET to encrypt the data in the hidden field. Then a malicious user would have to decrypt to find a legal GUID, generate some trial GUIDs, encrypt, guess an IP address, and post the data back to my server. And at that point, all they have managed to do is to have my server send some confused response to the IP address of the real user with that session ID. No biggie. Besides, my Web site is not for a bank or trading Bitcoins, etc.!

For "crazy", it can look that way to others. To me I just saw an unsolved problem, turned it into a math problem, got a math solution, and wrote some code. I believe that the problem is important and my solution good. We will see, at least see if users like my solution.

I tend to like doing: sha256(sha256($password) . $ip)

This encrypts the password and makes the cookie only usable by the IP it was set for. Then to verify the cookie, since I already store sha256 of the password, it's trivial to do, without having to store an additional token for persistence. Of course you can replace sha256 with your fav hashing function.

> since I already store sha256

Storing SHA-256 of passwords is a bad idea. You should use either scrypt, bcrypt, or at the very least PBDKF2 (with a large number of rounds).

For a signed token round tripped to the client you should use an HMAC[1].

[1]: http://en.m.wikipedia.org/wiki/Hash-based_message_authentica...

I have two ISPs at home. What happens when the next day my router decides to use the other IP address?

What if the user gets their session hijacked when they're on a static or long-term dynamic IP? If the user were to relogin, their session ID would be the same. I think it would be beneficial to at least store a salt associated with their session and regenerate the salt when authenticating.

sha256($password . $ip . $random_salt)

You could also regenerate the salt periodically.

No. None of these. Please stop. The system is entirely broken. No amount of tweaking is going to fix it.

In most parts of the world, users have dynamic IP's that change every time they connect to their ISP (often multiple times/day in the case of home users)

I'm not sure why you'd ever want to put something password-derived in a cookie--even if its sha256'd 8 dozen times, then scrypted, etc.

There's just no value in using the password as a basis. Why not instead just generate a random, unique token?

Remembering the username can be pretty useful, especially if you find yourself in a place where you common usernames are taken on other sites and you have to use different usernames in different places.

Being reminded of a username could also be helpful if you have more than one username at a given site - for instance, personal and business accounts for a bank or credit card site, per-project accounts for a contractor, etc.

Not to mention the sheer convenience of one less thing to type--especially helpful if you are on a mobile device, where typing is still a pain vs. on a full keyboard.

Or you can do this:

Client generates random value, puts it into cookie and passes it to server along with valid credentials. Server remebers it for this user and whenever at later time it sees this value in the cookie it logs in this user. Server forgets this value when this user logs out and after some time.

If client does't have javascript or you don't trust randomness it can generate you can create this value on server and make the server set this cookie when it responds to login.

What is the point of generating this value on the client? The second scenario you describe seems to work the same, minus the need to generate values on the client, with the same outcome.

I guess no point. Pick the one that you feel better with.

Also the server should keep it hmac-ed in case db silently leaked.

There's a lot of misinformation in this thread, so I'm going to describe best practice for this sort of thing.

  1. The client authenticates, and asks to be remembered.
  2. The server generates a cryptographically random token of at least 128 bits in length. This token is *never* directly stored server-side.
  3. The hash (or, possibly, a slow-hash) of the token is saved in the database along with a time of expiry
  4. The non-hashed token is sent to the client as a cookie completely independent of the session cookie

When the client visits the website:

  1. The server checks for the presence of the remember-me cookie
  2. If the cookie is set, it is hashed and this hash is searched for in the database (and filtered to tokens with expiry times after now)
  3. If the user is successfully logged in, the old token is deleted from the database. A new token is generated and sent to the client by the previously-described process

Now for a little bit of explanation.

This process completely isolates the authentication token from the session. If a token is somehow intercepted or discovered by an attacker, it is only usable for a single session. Such tokens should never be allowed to serve as permanent entry points into a user's account.

The token is never directly stored server-side. This confers two benefits. First, these tokens should be considered password-equivalent — if someone manages to steal your table of remember-me tokens, they would be completely unable to use them to log into a user's account. Second, this prevents timing attacks; if you simply looked for matching tokens in your database, an attacker can (shockingly simply) use timing information to guess a user's remember-me token in O(n) time.

I don't see why you have to store the hash of the token. You know the key you used for HMACing, why can't you just check that the cookie contains a valid hash of the rest of its data (which is as extensive as you need: session id, expiry, IP, whatever)? To prevent the use of old sessions just make the session id a counter. A single session counter per user is probably less hassle (and more useful) to store than the whole hash.

To avoid the problem of storing the token's hash, you have introduced:

  * parsing a structured cookie, vs a meaningless 128-bit string
  * securely storing and managing a secret HMAC key (which may be used to forge or modify credentials)
  * securely verifying an HMAC using a constant-time string comparision

So what, exactly, is the benefit of your proposal over mine? You've removed the (useful) distinction between an authentication token and the user's session. You've introduced significant additional complication and moving components. And you've increased the attack surface for security vulnerabilities. For what?

Doing all that to store a 32-bit integer is somehow less hassle than hashing and storing a 128-bit string?

you have introduced: parsing a structured cookie, vs a meaningless 128-bit string

You're going to need some logic for dealing with session, user, etc. metadata. Stipulating that parsing metadata out of a cookie is in some sense harder than reading it from a DB, keeping it in the database requires more storage (in addition to the 128-bit hash), more syncing among the various masters and slaves, and higher security. (An attacker who gets a DB dump can't see account metadata that isn't stored in the DB. Sure I'm storing a counter but what can he do with that?)

securely storing and managing a secret HMAC key

How do you run an online service without doing key management? Whatever techniques you use for your other keys, use them for this key too. It isn't as if we need a separate key for each user.

You've removed the (useful) distinction between an authentication token and the user's session.

I don't understand this claim. If I want to generate a new token for the current session, what's stopping me?

And you've increased the attack surface for security vulnerabilities.

Because HMAC collisions are easy now? I've moved some things around, but I'd say there is a tradeoff between the CAP costs of the "store everything" technique you propose and the slightly more involved validation process I describe. It's cool if you prefer your method for your situation, but can't we admit that "I'm going to describe best practice" was a bit overstated?

> Because HMAC collisions are easy now?

When you get the details wrong? Relatively so. And even experienced people get this shit wrong more often than they get it right. Rails had this vulnerability. Google's KeyCzar had this vulnerability. The average, non-crypto-enthusiast has effectively zero chance of getting this right out of the box.

Comparing server-computed hashes is inherently safer and less error-prone than comparing client-controlled HMACs.

  * https://github.com/rails/rails/commit/5e6dab8b34152bc48c89032d20e5bda1511e28fb
  * http://rdist.root.org/2009/05/28/timing-attack-in-google-keyczar-library/

> but can't we admit that "I'm going to describe best practice" was a bit overstated?

Not when it is accepted as best practice.

  * http://fishbowl.pastiche.org/2004/01/19/persistent_login_cookie_best_practice/
  * http://stackoverflow.com/questions/549/the-definitive-guide-to-forms-based-website-authentication

The only difference between my description is I do not store the username with the cookie; it is unnecessary and confers no discernable security benefit. I also avoid associating multiple remember-me tokens with a user, so an attacker who intercepts one must use it before the user next logs in.

Is your approach wrong? No. But it is more difficult for the average programmer to do securely and correctly, and confers few (if any) real benefits.

Thanks for the thoughtful responses!

Can someone explain how the example of the JSON encoded cookie could be used for XSS?

> What this means is that client script can access those cookies which means that if you can get a piece of XSS onto the site – just as you can on the Aussie Farmers site – you can steal cookies containing passwords if you can get a customer to load the XSS payload

The author is saying that, putting the password in a cookie means that, if there is an XSS vulnerability somewhere, it can be trivially used to steal usernames and passwords, since the XSS can be used to inject code on the site that can grab the values from document.cookie and append them to a hidden iframe.

It's really unclear, but I think it's a loose reference to an older hack that involves patching javascript's Array to steal data when parsing JSON. The references to 'old array' and 'new array' in the screenshot seem to be the tipoff. This is my best guess at least.

See http://flask.pocoo.org/docs/security/#json-security for info. Most modern browsers don't let you patch Array like this anymore.

Also, that's more a csrf issue so, I could be totally wrong.

I don't know if it's what Troy was implying, but JSON as a cookie payload along with the seemingly ubiquitous "parsing JSON using YAML and the YAML parser executing it as arbitrary code" vulnerability seems like it'd at least be worth attempting to attack there.

(Probably not what Troy was thinking, since that'd result in server-side SQLi or arbitrary code execution problems, not client-side ones like XSS. Also, I thnik Troy is more a .Net/asp kind of guy, not a Ruby-ist - which might make that YAML parser issue in Ruby/Rails not something he's concerned about)

I'm not sure either. I hope someone can enlighten.

Reply on news.ycombinator.com