HTML injection into someone else's website?

Question

I've got a product that embeds into websites similarly to Paypal (customers add my button to their website, users click on this button and once the service is complete I redirect them back to the original website).

I'd like to demo my technology to customers without actually modifying their live website. To that end, is it possible to configure http://stackoverflow.myserver.com/ so it mirrors http://www.stackoverflow.com/ while seamlessly injecting my button?

Meaning, I want to demo the experience of using my button on the live website without actually re-hosting the customer's database on my server.

I know there are security concerns here, so feel free to mention them so long as we meet the requirements. I do not need to demo this for website that uses HTTPS.

More specifically, I would like to demonstrate the idea of financial bounties on Stackoverflow questions by injecting a Paypal button into the page. How would I demo this off http://stackoverflow.myserver.com/ without modifying https://stackoverflow.com/?

REQUEST TO REOPEN: I have reworded the question to be more specific per your request. If you still believe it is too broad, please help me understand your reasoning by posting a comment below.

UPDATE: I posted a follow-up challenge at How to rewrite URLs referenced by Javascript code?

UPDATE2: I discarded the idea of bookmarklets and Greasemonkey because they require customer-side installation/modification. We need to make the process as seamless as possible, otherwise many of get turned off by the process and won't let us pitch.

Answer 1

I would suggest to create a proxy using a HTTP handler.

In the ProcessRequest you can do a HttpWebRequest to get the content on the other side, alter it and return the adjusted html to the browser. You can rewrite the urls inside to allow the loading of images, etc from the original source.

public void ProcessRequest(HttpContext context)
{
  // get the content using HttpWebRequest
  string html = ...

  // alter it

  // write back the adjusted html
  context.Response.Write(html);
}

Answer 2

If you're demoing on the client-side and looking to just hack it in quickly, you could pull it off with some jQuery. I slapped the button after the SO logo just for a demo. You could type this into your console:

$('head').append('<script src="https://www.paypalobjects.com/js/external/dg.js" type="text/javascript"></script>')
$('#hlogo').append('<form action="https://www.sandbox.paypal.com/webapps/adaptivepayment/flow/pay" target="PPDGFrame" class="standard"><label for="buy">Buy Now:</label><input type="image" id="submitBtn" value="Pay with PayPal" src="https://www.paypalobjects.com/en_US/i/btn/btn_paynowCC_LG.gif"><input id="type" type="hidden" name="expType" value="light"><input id="paykey" type="hidden" name="paykey" value="insert_pay_key">')
var embeddedPPFlow = new PAYPAL.apps.DGFlow({trigger: 'submitBtn'});

Now, I'm not sure if I did something wrong or not because I got this error on the last part:

Expected 'none' or URL but found 'alpha('.  Error in parsing value for 'filter'.  Declaration dropped.

But at any rate if you are demoing you could just do this, maybe as a plan B. (You could also write a userscript for this so you don't have to open the console, I guess?)

Answer 3

After playing with this for a very long time I ended up doing the following:

1. Rewrite the HTML and JS files on the fly. All other resources are hosted by the original website.
2. For HTML files, inject a <base> tag, pointing to the website being redirected. This will cause the browser to automatically redirect relative links (in the HTML file, CSS files, and even Flash!) to the original website.
3. For the JS files, apply a regular expression to patch specific sections of code that point to the wrote URL. I load up the redirected page in a browser, look for broken links, and figure out which section of JS needs to be patched to correct the problem.

This sounds a lot harder than it actually is. On average, patching each page takes less than 5 minutes of work.

The big discovery was the <base> tag! It corrected the vast majority of links on my behalf.