Should I use GET or POST when requesting sensitive data?

Question

Should I use GET or POST for retrieving sensitive data, given that:

• The response will contain sensitive data.
• There are side-eff

Answer 1

If the data is highly sensitive, consider using a POST. It's easy to issue a GET request without much thought - for instance if someone is looking at log file while logged in to the application with appropriate permissions and they click a link, they would generate a get request.

You can instead ask the client to construct a document you could call a 'sensitive data access request'. Potentially you could require a field with the reason for looking at this data. The server could receive this request and send the sensitive data in response.

Depending on the type of client application the users have access to, it might be much easier to demonstrate intentionality if a user sends a POST request than if they send a GET request.

Answer 2

In addition to Cássio Mazzochi Molin's excellent answer, you should use HTTPS but you should (generally) use:

• GET for retrieving sensitive data.
• POST for sending sensitive data.

The reason to use GET when retrieving is that the action does not have side-effects, therefore there is no reason to use POST. The only previously applicable reason to use POST was when retrieving JSON via AJAX, because old browsers had bugs meaning that another domain that the user had open in their browser could steal the data from the JSON using a <script> tag (JSON Hijacking). Disallowing GET prevented this attack because <script src="..."> always uses the GET method. See this answer. Note that using POST here means you should disable GET server-side for this method.

The reason to use POST for sending sensitive data is that it prevents data leakage via the query string (although another way would be to use GET with custom headers set, although POST makes much more sense). The reason is that query string data in the URL is logged by proxy servers, by server logs as default, and can also be stored in browser history, making it not a great place to transmit personal or otherwise sensitive details. Note that during transit over HTTPS they would be encrypted, it is just that they can leak from the encrypted state into other non-encrypted or non-controlled locations. Of course, going back to RFC 7231, if you're making changes based on this sent sensitive data, POST is the better idea as it'll prevent the browser accidentally sending it again in most cases.

One more reason to use POST is that modern browsers don't appear to cache the results of POST requests by default. However, this should not be relied upon. It is much better to set Cache-control: no-store header in your response either way, any time that sensitive data is output.

Answer 3

Using POST would only make sense as part of an application designed to prevent a user from repudiating receipt of the response. I don't know of any such schemes, and I wouldn't hazard to devise one off the top of my head.

The intent of the user's request isn't to create an audit log entry; the intent is to get the response. The accountability log is a side effect, but it's hidden from the user, so POST is not required.

In other words, you can't hold users accountable for getting the data, because you can't prove they received it. But knowing who requested the data might help an investigation, so recording it as a side effect is still useful.

Answer 4

A step back

First of all, the RFC 2616 is obsolete. Hence, it shouldn't be used as a reference anymore.

Below you'll find the current references for the HTTP/1.1 protocol:

• RFC 7230: Message Syntax and Routing
• RFC 7231: Semantics and Content
• RFC 7232: Conditional Requests
• RFC 7233: Range Requests
• RFC 7234: Caching
• RFC 7235: Authentication

The safe property

Have a look at what the RFC 7231 says about safe methods:

4.2.1. Safe Methods

Request methods are considered "safe" if their defined semantics are essentially read-only; i.e., the client does not request, and does not expect, any state change on the origin server as a result of applying a safe method to a target resource. [...]

This definition of safe methods does not prevent an implementation from including behavior that is potentially harmful, that is not entirely read-only, or that causes side effects while invoking a safe method. What is important, however, is that the client did not request that additional behavior and cannot be held accountable for it. For example, most servers append request information to access log files at the completion of every response, regardless of the method, and that is considered safe even though the log storage might become full and crash the server. [...]

Of the request methods defined by this specification, the GET, HEAD, OPTIONS, and TRACE methods are defined to be safe. [...]

In the context of HTTP methods, safe is not related to security and, in a similar way, safe is not about how you deal with sensitive data. Safe means read-only.

As stated above, the use of safe methods do not prevent you from performing operations that are not read-only, such as logging the request to a file. However, this operations should be transparent for the client.

Which method should you use?

It depends on the operation you are performing. In REST APIs, the POST method is frequently used to create resources while the GET method is frequently used to request a representation of a resource.

And how about security and sensitive data?

If you want to ensure security when sending sensitive data over the wire, use HTTPS and don't expose sensitive data (such as passwords) in the URL.

Answer 5

I would suggest using POST, not for any real technical reason, like side-effects, but rather because servers are often configured to watch POST calls more, and many off the shelf security modules consider POST as the place where the action happens.

Thats not really a great technical reason, but I would be interested in seeing what others think.

Answer 6

You should use GET for retrieving information from the server.

Auditing and logging would not be considered side-effects since they are transparent to the client.

Securing the response data can be done using SSL and "Cache-control: no-store". Once the sensitive data gets to the client, there's no way to stop them from doing whatever they want with it.