I am after some advice regarding use of GUIDs from the security perspective. I have developed an ASP.Net application. It provides the user with access to some material, such as documents and photos, that aren't on the web server. These are stored on a file server. I have a 'GetResource.aspx' page which takes the ID of the resource, opens it using System.IO.FileInfo writes it to the response stream and returns it.
So, GetResource.aspx?id=123 would return, say, a picture that the user has access to. Of course, the user could manually enter the URL as GetResource.aspx?id=456 in which case the picture / document etc with that ID would be returned and it may not be one they have permission to access.
So clearly using an integer ID is not adequate. Would using a GUID as the ID provide enough 'randomness' that I could reliably assume the user could never manually enter "GetResource.aspx?guid={A guessed guid}" and ever expect to access a valid resource, including if using a script that made many random guesses per second?
Or, is there no substitute to determining the ID of the user from a Session variable, determining he does actually have access to the requested resource and only then returning it (Which as I write this I'm more and more convinced is the case!).
Thanks
There is certainly no substitute to authenticating the user and seeing if they are authorized to access the resource. What you are proposing here is a method of making it harder for a user to hit on a valid id for a document they are not authorized to view (either by mistake or on purpose).
A GUID is certainly large enough that you would never get "accidental" valid ids in practice. That makes a GUID without authorization checks a system that works great as long as noone is actively trying to break it. On the other hand, authorization checking is a system that would work great even in the presence of active attackers (of course this depends on what the attackers can manage to do).
You should choose between the two approaches depending on the nature of your application (is it public? are the users known and accountable for their actions? how bad would a "security breach" be?).
You should be determining if the user is authorised before blindly serving it if it is protected content.
The GUID does help to some extent, it makes guessing URLs harder, so I'd still recommend using them. But URLs can still be shared (even accidentally). If you are just going to serve up the content anyway regardless of who makes the request then it is of little real protection.
If you think that content is restricted one and having some personal data then you should go with username and password thing.
Related
I'm developing a web app and I'm wondering if exposing UserIds to the clients is potentially posing a security vulnerability. (By UserId I refer to the Id of an Identity user object created by the Identity Framework and used as a PK of the users table.)
To give some context or an example: in my app, I need to distinguish between content that is posted by the signed in user and the content that was posted by others. In a naive approach, I would just compare the UserId of the content to the UserId of the currently authenticated user. But that would mean that the client sees the Ids of all involved users.
My gut feeling tells me that this is not a good idea but I couldn't pinpoint to why. So, I wonder if there's a guideline around this question. Maybe it's just the general rule to reduce the surface of knowledge that is exposed to a minimum.
If this is the case, how should I proceed? Would hashing of the UserId help to solve the problem or are there better approaches around?
EDIT
The example I made is not the best because this problem can easily be solved on the back-end by comparing user IDs there and then sending the contents to the client already tagged as "mine" or "by others". But still, the general question remains.
If user IDs are themselves sensitive data,for example, your primary keys for some reason happen to be social security numbers, that'll definitely be a security and privacy liability. If your user IDs are just auto-increment numbers though, it should be fine.
It is always best to expose a unique identifier other than the primary key outside your system. It gives you more flexibility in resolving data mix-ups, dealing with data migration issues, and in otherwise future-proofing your system.
If UIDs are just identifiers for users. Knowing a user's UID does not grant you any permissions that are associated with that user. Sharing the UID in URLs is about as safe as sharing your username on Github, or your unique ID on Stack Overflow.
Stack Overflow displays user IDs in their URLs in order to make user profile lookups work: https://stackoverflow.com/users/10158551/xing-zou
Anyway, it is up to your design and you need to consider more than we could.
Refer to Should I expose a user ID to public?
I would put this in the low risk category. Exposing user ids does increase the risk surface as you have correctly identified.
For you application, it sounds like you require an identifier for the post and an indication of whether that post has been made by the user or by someone else.
You could structure that in the following way to avoid the low risk exposure:
GET /posts
{
"postId": "AJDIWC",
"isAuthor": true,
"content": "This is a post by the user."
},
{
"postId": "LISHWE",
"isAuthor": false,
"content": "This is a post by another user."
}
EDIT
To answer your question following the edit, yes, it is best practice to avoid exposing identifiers from other users unless the authorisation boundary allows it.
For example, there is no issue having an admin user of an RBAC protected application see the unique identifiers of users within their security boundary.
It's usually not a good idea to expose the user id at the client-side. If you need any information, username or a unique number is better than the exact user id in the database. This is generally not an issue if your application has protections against SQL injection attacks. But in case there is any vulnerability in loopholes, if anyone knows the user id for a user, they can inject SQL scripts for that user.
A better idea is to issue an access token to your clients with claims inside that. The access token will perform authentication for you on the server-side.
i've read all the tutorials and questions asked about the subject but they contradict one another
claims(userclaims and roleclaims) are serialised into a cookie (along with custom identity user properties you specify by overriding the principle factory) but they don't get queried, they're stored in a cookie, which mean that the more claims a user have, the more data will have to round trip between the server and browser
a custom identity user property by default don't get serialised but get queried from the database, so every time you need that data it query it from the database which is more work to do on the database on each request if you frequently query for it
so which is more efficient and which is safer
for instance
IsAdmin should be a role/claim? but then if someone stole the cookie, nah nah, the cookie already contains userid/username/securitystamp, so even if it's a property, the userid on the stolen cookie would query on the custom identity user property, or is there something that will prevent this cookie from working when stolen ?
another instance
if i've 20 property for the user (first name, last name, address 1, address 2, postal code, whatever the case may be), should i simply have the user wait a bit for a bigger slower cookie to be send back and forth or should i do all the work from the db using custom identity user
but then, if i remove or add a claim to the user, would it be updated on the next request if it doesn't get queried or is the security stamp validate that this cookie is still valid ?
cause at the Task AddClaimsAsync of the userstore of efcore it only add the claim to the dbset
i apologize i know this is many questions to ask but the resources on the matter are not that good and one can easily get lost reading the identity source
Rule of thumb - put frequently added items as a claim, everything else can live in DB and be queried on demand. I.e. address 1, address 2 can't be required on every request, so keep them in the DB.
On the other hand IsAdmin (should be a role anyway) I can imagine will be checked on every request, so it should be in the cookie without having to query the db.
If you afraid of your cookies getting stolen, don't show them to anyone! set up SecurityStampValidator to do frequent checks - like every 5 minutes. This basically updates the cookie with the fresh information from your database and changes the cookie. So even if the cookie is stolen, it will only work for 5 minutes.
I don't think the two statements contradict, it depends on your configuration. The second statement mentions "by default".
You do not need to store all information in claims nor do you need all the information all the time. If you want profile information, call the server once and store the information on the client, or get it when needed only.
The same counts for authorization, in case you want to show / hide elements based on permissions. This may include a tag called "IsAdmin". Authorization should be kept close to the resource.
If your client wants to refresh the information, just call the server. Claims are not updated during each request. In general, the user has to log out and log in again. So Claims are not flexible and therefor not really suitable for properties that can change (frequently).
As for safety, it doesn't really matter that the client can alter the information, it is for display only. It doesn't change the permission in the backend.
You can consider to add something like a display name to the properties, if you are showing that in every page. You can also consider to implement caching to limit database calls. In the end it really depends on your requirements.
As for stolen cookie, you'll need to implement additional security to your server to detect suspicious behaviour. You may want to include the ip address as claim. As for the admin, add security, e.g. filter by ip address and / or use an additional code which was send by email.
When a customer signs up for a site, we want to let them know whether a username/email is available for use.
We have a httphandler that serves the purpose of feeding a jquery script that showsthe customer whether or not their desired username/email is available.
The question is:
The service can clearly be seen being called when you view the request in fiddler.
It shows /emlhandler.asmx?name=xxxxxxxxxxx#yyy.com
From the handler, a simple 0 or 1 is returned to indicate whether or not the name/address is available.
My concern is that this seems like a major security issue that would be very easy for an inexperienced person to exploit to discover all the users on the site.
So friends, how do you protect your site info and still allow the ajax callback to provide a great user experience?
Thanks.
Heath
You are being slightly paranoid here. Every site that allows user registration has to do something similar. One thing you can reasonably do is add a slight delay (maybe 2 or 3 seconds) to the handler's processing in order to reduce the likelihood or ease of a brute-force attack. Frankly, I don't think anyone would bother.
Another option is just to ignore repeated emails and send a verification email before a user's registration actually becomes active. If a new user attempts to use an existing email, the original email owner receives the verification and can cancel or ignore it. But I don't recommend this.
I'd say the vast majority of the sites I've used will just immediately say "this email address is already registered... did you forget your password?" Just knowing an email address is already in use on a given site does not in itself represent a security breach.
One possible solution would be to only enable POST requests for that method.
And since you cannot invoke services from JavaScript from another domain (XSS - Cross-site Scripting) without your authorization, then you would be protected.
However this technique would prevent malicious users from calling your web service to discover user names but this wouldn't prevent the user to automate a process to simulate user entering data in a text box to force a call to the service, in that case, perhaps you could allow just a number of requests per user in an X amount of time.
You could keep track of the number of attempts using the Session object from your web service
Another approac would be to add a re Captcha to your site, however this would decrease the level of responsiveness if you used to allow your users to capture a user name and as soon as they write you call your service. Implementing would require your users to write the auto-generated captcha in order to submit your data
I'm using Forms authentication in ASP.NET MVC website and I store user account login name in AuthCookie like this:
FormsAuthentication.SetAuthCookie(account.Login, false);
I want to ask if there is a possibility that user on client side will somehow manage to change his login name in AuthCookie and thus he will be for example impersonated as someone with higher privileges and authorized to do more actions than he is normally supposed to have. Also is it better to save in this cookie user account login name or user account ID number?
The cookie will be encrypted and decrypted on the server side, so unless the user can crack the encryption key, he or she won't be able to do this.
As long as the information you store uniquely identifies your user, the choice as to what that information is is entirely down to the requirements of the particular application.
No it is not possible (well, in theory maybe but it's not feasible in practice). The value of the authentication cookie is encrypted so the user can not tamper with it. It is a good idea to store the (unique) login name in the authentication cookie, because when the IIdentity object (HttpContext.Current.User) is restored, the value that you passed to SetAuthCookie is used for the Name property of the IIdentity. The Name property will be shown if you use the LoginStatusControl, for example, so it's a good idea that the value of the Name property makes sense to the user.
Cookies are encrypted so chances for that a quite slim. But still.
More than one property approach
If you'd like to make your security even tighter you could save username as well as user ID or some other data that can't be guessed from the username. The combination of these makes it safer because if you can guess one it harder to guess others and use the correct combination of them. Ie. If you guess other user's email/username it's a bit harder to guess the same user's ID, because they're not related. So the more unrelated properties you combine the more steps it takes to get the right combination.
Logon security token approach
You could use an alternative approach described in this scenario:
User logs in.
Generate a random security logon token that can be of random length with some minimum length defined and save it against user in the data store. This is probably not a problem while it's quite common that other data is stored at logons as well like LastLogonDate info.
Use this token and save it in the cookie instead of usernames or other info.
If user logs-out, clear logon security token from the data store
User logs in again... go back to 1 and create a new token again and use it in this session.
This way it will make it safer on the long run, because this information will change on each login and if user does manually logout you can always clear that token from the store, so it won't at all be possible to inject someone else's identity. This does make it a but more complicated to use permanent cookies though but it can still be done.
This approach is not bullet proof but it provides additional security level that prevents the same attack over and over again when one account has been compromised. And also when one account is compromised it doesn't mean that others can be as well. If your security tokens are long enough it would be much harder to start a brute force attack on your site and while this kind of attack would be executed security tokens will change so it is definitely safer.
I think, in almost all cases user preference data may be stored in a cookie with (almost) equally good results as when the User Profile API is used. Disadvantages of using cookies (for authenticated users) seem to be that a cookie can be deleted or time-out, in which case the user preference data will be lost. For anonymous users, if the preferences data needs to be persisted across sessions then a cookie will have to be used even when User Profiles are used.
So what are some of the biggest advantages/disadvanges of using either User Profiles or cookies for storing user preferences?
One of the benefits of registering on a site is that it remembers my preferences - if you're storing that information in a cookie on my machine instead of on your server then when I log into your site from another computer, I've got to set all my preferences up again - from a usability point of view, this is fairly bad.
For an anonymous user, storing the prefs in a cookie may seem fairly sensible - you don't know who they are, or whether they will comeback, and as you state, you can't work out from one session to the next who they are - however you'd probably be better off storing some sort of token in the cookie and mapping that to a preferences store on the server.
Also, I've noticed different browsers have different implementations for cookies - for example IE can now receive 50 cookies from one domain (up from the original 20), but it is still limited to a total of 4096 bytes for the entire cookie collection (and previous) - other browsers will support 4KB per cookie, rather than per domain.
Another disadvantage to holding all the preference data in cookies is that all of that data will have to be sent in every request from the client and in any response from the server whenever a change to the data is made. Whilst this may seem like a minor point in the age of broadband it is still an additional overhead. Using the Profiles API meands that the data is held at the server and only a session identification cookie needs to be sent by the browser.
Also, as you stated, for anonymous users if cookies are deleted then the user preferences held in the Profiles DB will no longer be accessible. However this will not be the case with registered users of your website. If they remove their cookies the server will still be able to retrieve their user preferences the next time they log in.
Cookies are limited in maximum length and they are using an implementation beyond of your control (after all, they are a feature of your visitors browser). Personally, I dislike relying on unknown third-party implementations I don't have any control over and if I have to, I'm trying to use it in the simplest way possible.
So from where I'm coming from, I would always store the user data on the server and just pass around a cookie pointing to that information.
Aside of not trusting the browser with a potentially big chunk of data (which may be lost, incorrectly stored or not stored at all depending on not only the browser but also, say, some antivirus application or whatever), this has various other advantages:
You are hiding your implementation from the user: If you store the data in the cookie, it's visible for anybody and can be analyzed or modified at will. This can even lead to users changing cookies to there liking and thus force you into keeping stuff around you probably want to get rid of just because some users are depending on your particular implementation at any time.
As cookies are stored in plain text, on shared machines, everybody can no longer easily see all the settings the previous user made, nor change them at will.
But the most important point remains the disconnect from not-quite-working browser implementations (just storing small tokens is the common, tested use-case)
Don't forget that one of the biggest disadvantages of using cookies is that they can be copied, so its dangerous to store authentication info on them.
I'm not familiar with User Profile API but I'm guessing it stores the information on the server(?). If thats the case then you could have a problem if you have to many users.
Overall maybe the best solution is to use User Profile if it guarantees the persistence of the information.
Keep in mind that its possible to write a ProfileProvider that persists user data in a cookie, so you can have the best of both worlds if you determine the state you want to persist is appropriate for cookies (size, security, etc).
Actually, you do not need to persist preference data in cookies for anonymous users when using the ASP.NET Profile Provider. Simply store the current UserID (which is some horrible looking session-related string) in a cookie. This becomes the previous UserID on subsequent visits, and then you can just grab the old Profile information and migrate it to the current Profile, or even authenticate them as that old anonymous Profile.