sudofox's journal

Austin Burk's journal, where I share little snippets of my writing, code, and dreams.

Google is Scanning for (and Crawling) URLs in Your Private YouTube Videos

I was recently uploading an unlisted video to YouTube to demonstrate an XSS vulnerability I stumbled across which I was responsibly disclosing. Part of this involved showing the URL of the script which had been run. After uploading it to YouTube and submitting the vulnerability disclosure, I decided to double-check that nobody had visited the page I was testing on before I had removed the link. As it turns out, somebody had: YouTube.

66.249.90.79 - - [12/Dec/2018:14:23:40 -0500] "GET /js/redacted_1.js HTTP/1.1" 200 125 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.81 - - [12/Dec/2018:14:23:42 -0500] "GET /js/redacted_1.js HTTP/1.1" 200 125 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:15:24:21 -0500] "GET /redacted_subfolder/redacted_2.png HTTP/1.1" 200 4605 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.81 - - [12/Dec/2018:15:24:22 -0500] "GET /redacted_subfolder/redacted_3.png HTTP/1.1" 200 5102 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:15:24:23 -0500] "GET /js/redacted_4.js HTTP/1.1" 200 137 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:15:24:24 -0500] "GET /redacted_subfolder/redacted_2.png HTTP/1.1" 200 4605 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:15:24:26 -0500] "GET /redacted_subfolder/redacted_3.png HTTP/1.1" 200 5102 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:15:24:26 -0500] "GET /js/redacted_4.js HTTP/1.1" 200 137 "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"

I was rather alarmed to see this, as I didn't imagine the links were up long enough to be crawled by Google. It was then that I realized that during the video, those URLs were visible in the address bar. It seemed that YouTube had run OCR (optical character recognition) across my entire video and decided to crawl the links within. But how could I be sure that this was not just a mistake on my part?

Time for an Experiment

I recorded a new video of me accessing a URL that does not exist for the very first time.

Here is the video that I uploaded:

I started another screen recording of me uploading the video, and watching the access logs. A few minutes later, Google took the bait, and sent two requests to the URL:

66.249.90.81 - - [12/Dec/2018:18:42:02 -0500] "GET /nonexistent/url.js HTTP/1.1" 404 - "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"
66.249.90.79 - - [12/Dec/2018:18:42:04 -0500] "GET /nonexistent/url.js HTTP/1.1" 404 - "-" "Mozilla/5.0 (compatible; Google-Youtube-Links)"

Hook, line, and sinker! I recorded me uploading the video and watching my access logs live (the accesses are around the 5:50 mark):

Why is this concerning?

f:id:austinburk:20181213091541p:plain

The purpose for which I uploaded the video was to report a vulnerability. I uploaded it unlisted, so far all intents and purposes, it was meant to remain private. However, our friend Google-Youtube-Links scanned it for an unknown purpose and sent several requests to that URL. A second test as a fully private (not just unlisted) video revealed the same result.

By uploading the videos as unlisted or private, I have the expectation that nobody will see the video or the contained within except for me, or for the people who I explicitly share the links with.

Let's propose a scenario which is in a similar realm to what I was doing:

A security researcher has found a critical vulnerability in a site, and has crafted a URL that will trigger it, causing harmful effects to the website. (e.g a SQL injection vulnerability that will drop the database tables).

During the video, s/he makes mention that they will not visit the URL as it would cause trouble, but it is displayed so that the company they are responsibly disclosing to can remedy it. They upload it as unlisted to YouTube and submit their report. Five minutes later, Google-Youtube-Links comes along and sends two requests to the URL, triggering the SQL injection and rendering the site broken.

The Illusion of Privacy 

Here is Google's explanation of privacy settings:

What this does not include, however, is any mention that your video will be scanned for anything resembling URLs, and have these crawled.

What Google has to say about it

So what does Google have to say about this practice? Actually, nothing at all. Searching for the user-agent gives no relevant results save for one: a locked thread from a curious webmaster with no answer, back on March 27th, 2018. The thread was not resolved.

https://productforums.google.com/forum/?hl=en#!topic/webmasters/Ov_ODO8l2cU

This means that we are left with no explanation of why this is occuring, or disclosure that content uploaded as private to YouTube will be scanned with OCR and have any links within crawled by Google.

Honestly, I find this rather unsettling - especially for using private or unlisted YouTube videos as a way to quickly upload a video to disclose a vulnerability. I'm sure you can think of other scenarios in which this would be undesired, especially as we don't know why it's taking place or where those URLs will end up.

Let me know what you think of this development.

Questions or concerns?

If you have any questions or concerns, feel free to leave them below, or reach out to me at apburk@sudomemo.net. If you are a Hatena user, feel free to leave a star!