Better techniques than url parameter encryption

Question

I'm a programmer working on an application where the only choice/vs/deadline was to implement symmetric encryption on url parameter values. The data is insensitive in nature, but we needed to prevent sales agents from peeking on each other's leads. (Keys are generated on session creation and are cryptographically strong.) Sessions are expected to end frequently.

The role hierarchy was Manager--> Supervisor--> Agents. The data structures don't currently account for these roles in a way to strictly enforce who can see what. Getting this information from the database was NOT anywhere close to straightforward. (Recursive Database.)

I know that this technique is way down on the list as a defence against parameter manipulation. What would have been a better technique?

Constraints:
Role-based checking is not an option.

[Additional information] The urls built and sent to the client before I made any changes looked like:

https://www.example.com/agent/?producerId=12345

The specific threat surface here is parameter manipulation against ?agentId=12345. Agent ids are assigned uniquely to each agent. So if Agent A wants to look at Agent B's stats, he could have entered agentId=22222 in order to look at that agent's quotes and current sales statistics.

Again, Role-Based checking was not an option for me: I was unable to make changes to the database OR the persistence tier.

My solution was to use a session-created encryption key (using Java's KeyGenerator class) and encrypting the outbound urls sent to the client. So now, the url looks like:

https://www.example.com/agent/?producerId=<ciphertext>

Now, if someone tries agentId=22222, the server will decrypt what it thinks is ciphertext and will ultimately create an invalid character sequence.

(This leaves open the possibility that an existing agentId could be found, but quite unlikely that it would be relevant to the person performing the attack.

I will stress that this question isn't about optimal security (which would be role-based checking to ensure resource access) and about trying to squeeze some security in a grey area.

The parameter encryption solution here was recommended to me by one of our security guys. I got one takeaway I hadn't considered on this solution--broken urls--and will be using that as well as the maintenance issue created by this solution to argue for the time to enforce the access rules in a less stopgap fashion.

Answer 1

Good question! Thanks for elaborating on the threat you are trying to defend against. I have edited my answer accordingly.

Summary. Your primary defense should be access control. You need to limit which users can view which pages. Details below.

Access control in web applications. What you need to do is check that the user is authorized to access the data you're going to show on a page, before allowing them to see that data. This basically comes down to access control: you want controls that limit which users can view which data, based upon some authorization policy.

It sounds like you have a sequence of pages, one for each agent:

http://www.example.com/agent/?producerId=12345
http://www.example.com/agent/?producerId=12346
http://www.example.com/agent/?producerId=12347
...

where the producerIds (agentIds) are potentially guessable or predictable. You want to ensure that agent 12345 can view http://www.example.com/agent/?producerId=12345 but not any of the other pages. OK.

This is a bog-standard situation, and the bog-standard defense is: access control.

To implement access control, you code the web application so that each page checks whether the user is authorized to view that page before allowing the user to view that page. For instance, for the page listed above, the logic implementing that page would check the identity of the currently-logged in user. If the id of the logged-in user matches the producerId of the page parameter, then you show them the information. If the id does not match, you do not show them the information: if it is some other user, you show them an error page (with information about how to get access), or if the user has not logged in yet, you redirect them to a login page.

This won't break bookmarks. It does not require changes to the database, changes to the persistence layer, or role-based access control. It does require you to have a way to look up the identity of the currently logged-in user and associate that with their provider ID. Also, if you want to allow manager and supervisors to see the data for all other agents, then you need a way to look up the currently logged-in user and determine whether they are a manager or supervisor or not. If you want to allow only the agent's manager/supervisor to view their page (not all other managers/supervisors), then you need to have a way to determine the manager/supervisor of each agent. These are pretty basic, minimal requirements; it is hard to see how you could avoid them.

As @symbcbean properly points out, this is a very common error frequently found in web applications. A typical example might be a site that uses some guessable parameter value to identify a resource, and does not adequately authenticate the user. For instance, suppose orders are assigned a sequential order number:

https://www.example.com/show_order.php?id=1234
https://www.example.com/show_order.php?id=1235
https://www.example.com/show_order.php?id=1236
...

and suppose that anyone who knows the URL can view the order. That would be bad, because it means that anyone who knows (or guesses) the order number can view the order, even if they are not authorized to do so. This is one of OWASP's Top Ten web application security risks: Insecure Direct Object References. For more information, I recommend reading the resources available on OWASP. OWASP has lots of great resources on web application security.

Other comments. Others have suggested using SSL. While that will not prevent parameter tampering, it is a general good security practice that defends against other kinds of problems. Using SSL is straightforward: just configure your website to use https, instead of http (and ideally, enable HSTS and set the secure bit on all cookies).

Also, it is often better to avoid storing confidential information in URL parameters, all else being equal. You can store the confidential information in session state or in the database.

Answer 2

In short: Don't encrypt URL parameters, use a separate look-up.

Also, using HTTPS is basically non-negotiable if you desire any measure of web application security. It's mandatory in 2015. Get comfortable with TLS 1.1+.

---

What developers want to do

What developers should do instead

Answer 3

but we needed to prevent sales agents from peeking on each other's leads

This rather implies that the client is a browser - are you sending the key as cleartext at some point?

Polynomial is correct, you should be using SSL. That's not going to solve the problem of users typing in adjacent values to a URL looking something like:

https://www.example.com/show_order.php?id=1234
https://www.example.com/show_order.php?id=1235
https://www.example.com/show_order.php?id=1236
...

It's quite possible to generate an authentication token server-side based on the parameters which must be presented to validate the request. Ideally you would use a message authentication code (MAC) for this, but a hash would work too if you are careful. e.g. in PHP...

 print "<a href='show_order.php?id=" . $id . "&valid=" . md5($id . crypto_key()) . "'>...

Which is validated simply by:

if ($_GET['valid'] != md5($_GET['id'] . crypto_key()) {
   die('not authorized');
}

Here crypto_key() returns a static cryptographic key (generate it by pulling, say, 128 bits from /dev/urandom and storing it in the database).

But you still need to control access to the code which generates the URL.

Answer 4

Here is my solution

$id=1234;
$en_id = encrypString( $id);

and I create the url like

https://www.example.com/show_order.php?id=$en_id

the url will look like

https://www.example.com/show_order.php?id=9muEYh4lShFDeCnXqoNpxucs42Fuz5Nexq1IUGWYEffffe88yRbJu

and on the other side I decrypt

$en_id= decryptString($_GET['id']);

the functions for crypt and decrypt are

function encrypString($plaintext) {
         # --- ENCRYPTION ---

        $key = pack('H*', "bcb04b7e103a0cd8b54763051cef08bc55abe029fdebae5e1d417e2ffb2a00a3");//change this

        # show key size use either 16, 24 or 32 byte keys for AES-128, 192
        # and 256 respectively
        $key_size =  strlen($key);
        //echo "Key size: " . $key_size . "\n";


        # create a random IV to use with CBC encoding
        $iv_size = mcrypt_get_iv_size(MCRYPT_RIJNDAEL_128, MCRYPT_MODE_CBC);
        $iv = mcrypt_create_iv($iv_size, MCRYPT_DEV_URANDOM);

        # creates a cipher text compatible with AES (Rijndael block size = 128)
        # to keep the text confidential 
        # only suitable for encoded input that never ends with value 00h
        # (because of default zero padding)
        $ciphertext = mcrypt_encrypt(MCRYPT_RIJNDAEL_128, $key,
                                     $plaintext, MCRYPT_MODE_CBC, $iv);

        # prepend the IV for it to be available for decryption
        $ciphertext = $iv . $ciphertext;

        # encode the resulting cipher text so it can be represented by a string
        $ciphertext_base64 = base64_encode($ciphertext);

        return  rawurlencode($ciphertext_base64);//important rawurlencode for + symbol in url

    }


decryptString($ciphertext_base64) {
        # --- DECRYPTION ---

        $key = pack('H*', "bcb04b7e103a0cd8b54763051cef08bc55abe029fdebae5e1d417e2ffb2a00a3");//change this

        # show key size use either 16, 24 or 32 byte keys for AES-128, 192
        # and 256 respectively
        $key_size =  strlen($key);
        //echo "Key size: " . $key_size . "\n";

        # create a random IV to use with CBC encoding
        $iv_size = mcrypt_get_iv_size(MCRYPT_RIJNDAEL_128, MCRYPT_MODE_CBC);
        $iv = mcrypt_create_iv($iv_size, MCRYPT_DEV_URANDOM);

        $ciphertext_dec = base64_decode($ciphertext_base64);

        # retrieves the IV, iv_size should be created using mcrypt_get_iv_size()
        $iv_dec = substr($ciphertext_dec, 0, $iv_size);

        # retrieves the cipher text (everything except the $iv_size in the front)
        $ciphertext_dec = substr($ciphertext_dec, $iv_size);

        # may remove 00h valued characters from end of plain text
        $plaintext_dec = mcrypt_decrypt(MCRYPT_RIJNDAEL_128, $key,
                                    $ciphertext_dec, MCRYPT_MODE_CBC, $iv_dec);

        return rawurldecode($plaintext_dec);
    }

Answer 5

To prevent parameter tampering, I've always just sent along a hash with the plain text values.

e.g. take this url you want to secure

https://www.mysite.com/somepage?param1=abc&param2=xyz

On the server, your model would hash all the url values with a secret salt

string salt = "A1B2C3D4...";
string param1 = "abc";
string param2 = "xyz";
string hash = YourFavoriteHashingAlgorithm(param1 + param2 + salt);
// result hash = "Y83YMB38DX83YUHFIEIGKDHSEUG"

then you send this hash along with the other url values

https://www.mysite.com/somepage?param1=abc&param2=xyz&hash=Y83YMB38DX83YUHFIEIGKDHSEUG

Now, when you receive a request to this URL, you again take the parameters you are presented and hash them through the same algorithm. The hash you just generated should match the hash you are being presented, otherwise send them a 400 "Bad Request" result!.

The nice thing is your parameters are still readable by humans, and all your existing validation logic can remain the same.

Answer 6

Do not use user input

(because you camnot trust it)

This answer extends the accepted with what looks to me a significant simplification.

Now, from your description and from my best understanding, you have said that you want to prevent Sales Agent A (namely 12345) to peek into Sales Agent B's (namely 54321) data.

Simply, kill the agentId parameter from the query string and get it from the session

The URL becomes https://example.org/show_order.php

Internally, the application must extract the sales agent id from the principal stored within the session. I am excessively rusty in PHP so I will use pseudo code

SELECT * FROM sales where salesman_id = ?1;
[1 = getPrincipalSalesId()]

This query will simply ignore everything that comes from the client. It does not require modifications to the persistence layer. It doesn't even require implementing RBAC (role-based access control), but everything is about that function getPrincipalSalesId.

It's basically the same code you use to generate the URL, but this time you hammer that value into the query, making it implicit.