In this post, I consider the information security implications around using a secure IFRAME (which, for example, hosts a payment or checkout page) on an insecure container page.
- a "secure IFRAME" is one where its
srcattribute points to a domain over HTTPS, and any forms within the IFRAME submits to an HTTPS domain
- an "insecure container page" is a "container" because it hosts the IFRAME (the HTML code for the page will have an
<IFRAME>tag somewhere within it), and "insecure" because it was loaded over HTTP, not HTTPS.
Opinions in the wild
TL;DR Regardless of how secure the IFRAME is, the container site can be compromised and render said security moot.
Start typing "secure iframe" in Google, and note the suggestions:
The search results point to related questions asked on the Information Security Stack Exchange Q&A site, and a few blog posts by security experts have been written on the topic.
A non-exhaustive list of examples:
MITM attack is a "man in the middle" attack.
You go to your favourite coffee shop, and the following happens
- log on to their free Wifi
- frequent your favourite online shopping fix
- your requests to see products and pay for products go through many "hops" before it reaches the destination
- your browser connects to the Wifi access point
- the latter connects to the coffee shop's ISP via one hop or many
- the ISP connects to a shared exchange somewhere in Britain via one hop or many
- the exchange connects to an exchange somewhere in another country via one hop or many (and almost definitely via a long undersea cable)
- the overseas exchange connects to the shopping site's web server in a data centre somewhere (and all together now!) via one hops or many
Any of these "hops" can be compromised, but the typical scenario is for the attacker to host their version of "free Wifi" right from their laptop with a legitimate-sounding name like "FreeHub".
At this point the attacker's laptop is just one more "hop" you're totally oblivious to when you connect to "FreeHub", letting your surfing activity through uninterrupted.
However, since the attacker now has access to the data stream between you and the outside world, she can now see the data. The crucial differentiator here is HTTP VS HTTPS: if you access a site directly via HTTPS, the attacker will see unintelligible garbage. This is A Good Thing. If you access a site via HTTP, the attacker sees everything in the clear. Even if the website subsequently redirects to HTTPS, there's an opportunity here for the attacker to spoof a TLS certificate.
Since the attacker now has a clear stream of data to intercept, the attacker can also change what your web browser sees (not you, necessarily).
Example of a DIY MITM attack
I'll explore one example of exploiting an IFRAME (secure or not!) from an insecure container page. I'll take some shortcuts in the sense that I won't set up a compromised Wifi access point, or run software which injects malicious code into intercepted web traffic.
Clone this gist, or follow these steps:
Create this simple index file:
<p>IFRAME below</p> <iframe id="frame1" src="http://localhost:8000/i-dont-matter.html"></iframe>
Create this simple form which will be loaded by the IFRAME:
<form action="https://httpbin.org/post" method="POST"> <input name="secret" placeholder="Secret stuff here"> <input type="submit"> </form>
Host the code locally with Python:
python -m SimpleHTTPServer
The above the MITM attack and is what the attacker could do with the "in-the-clear data stream" we discussed before. The attacker just matched on a known token (like the form submit button) and injected custom code with which to do evil.
Submit the form, and see the side effect.
Note that the IFRAME's contents have been entirely modified. It was even made to look like the original content. The fact that the IFRAME's
src attribute previously loaded a secure site DOES NOT MATTER.
Consider yourself alerted.
What evil side effects are there?
alert you just saw is just for show, but consider this:
- the attacker sends the sensitive data you just submitted off to their domain to be harvested à la Tunisia
- the attacker changes the delivery address for the expensive item you just bought to the unused address down their street
Who cares about my site?
Your insecure site mightn't accept sensitive input like payment details or sell valuable items, but your vulnerable site might be frequented by someone who does.
An attacker might know that a high profile (read: hack-worthy) individual called Bob Millionaire frequents your insecure pet discussion website. Let's imagine for a second your forum software requires a login. Your insecure website has now become a target.
The attacker will now trivially sniff for Bob Millionaire's password on your insecure website, which she can replay against other more lucrative websites if we can presume that Bob Millionaire re-uses his passwords.
Even if you're just the third party who owns the domain the IFRAME loads, you have to strongly consider if you want to be associated with the bad press which comes with your client's website being exploited and having the IFRAME which was destined for your content being on the receiving end.
Can we get to a better place by mandating all container sites to enforce TLS? I think we can. But this will just be a better place, not the ultimate secure place.
There's a sliding scale, you see.
On the paranoid end, you can presume every device and every network is compromised. No amount of TLS will help you. You can trust your laptop as much as you do the web browser running on the old XP PC down the Internet cafe: you'll think twice about what you do online, and where you do it. (yes, a modified version of any web browser can show a green padlock for anything)
On the naive end, you can presume that you live in a free country, with devices built in secured facilities by known brands, and that the websites you frequent are hosted in secure data centres, managed by security veterans who can be trusted.
Choose your position on the scale, and accept the risk that's forthcoming.