More on Backdooring (or Not) WhatsApp

by Bruce Schneier on August 2, 2019

Yesterday, I blogged about a Facebook plan to backdoor WhatsApp by adding client-side scanning and filtering. It seems that I was wrong, and there are no such plans.

The only source for that post was a Forbes essay by Kalev Leetaru, which links to a previous Forbes essay by him, which links to a video presentation from a Facebook developers conference.

Leetaru extrapolated a lot out of very little. I watched the video (the relevant section is at the 23:00 mark), and it doesn’t talk about client-side scanning of messages. It doesn’t talk about messaging apps at all. It discusses using AI techniques to find bad content on Facebook, and the difficulties that arise from dynamic content:

So far, we have been keeping this fight [against bad actors and harmful content] on familiar grounds. And that is, we have been training our AI models on the server and making inferences on the server when all the data are flooding into our data centers.
While this works for most scenarios, it is not the ideal setup for some unique integrity challenges. URL masking is one such problem which is very hard to do. We have the traditional way of server-side inference. What is URL masking? Let us imagine that a user sees a link on the app and decides to click on it. When they click on it, Facebook actually logs the URL to crawl it at a later date. But…the publisher can dynamically change the content of the webpage to make it look more legitimate [to Facebook]. But then our users click on the same link, they see something completely different — oftentimes it is disturbing; oftentimes it violates our policy standards. Of course, this creates a bad experience for our community that we would like to avoid. This and similar integrity problems are best solved with AI on the device.

That might be true, but it also would hand whatever secret-AI sauce Facebook has to every one of its users to reverse engineer — which means it’s probably not going to happen. And it is a dumb idea, for reasons Steve Bellovin has pointed out.

Facebook’s first published response was a comment on the Hacker News website from a user named “wcathcart,” which Cardozo assures me is Will Cathcart, the vice president of WhatsApp. (I have no reason to doubt his identity, but surely there is a more official news channel that Facebook could have chosen to use if they wanted to.) Cathcart wrote:

Sponsorships Available

We haven’t added a backdoor to WhatsApp. The Forbes contributor referred to a technical talk about client side AI in general to conclude that we might do client side scanning of content on WhatsApp for anti-abuse purposes.
To be crystal clear, we have not done this, have zero plans to do so, and if we ever did it would be quite obvious and detectable that we had done it. We understand the serious concerns this type of approach would raise which is why we are opposed to it.

Facebook’s second published response was a comment on my original blog post, which has been confirmed to me by the WhatsApp people as authentic. It’s more of the same.

So, this was a false alarm. And, to be fair, Alec Muffet called foul on the first Forbes piece:

So, here’s my pre-emptive finger wag: Civil Society’s pack mentality can make us our own worst enemies. If we go around repeating one man’s Germanic conspiracy theory, we may doom ourselves to precisely what we fear. Instead, we should we must take steps to constructively demand what we actually want: End to End Encryption which is worthy of the name.

Blame accepted. But in general, this is the sort of thing we need to watch for. End-to-end encryption only secures data in transit. The data has to be in the clear on the device where it is created, and it has to be in the clear on the device where it is consumed. Those are the obvious places for an eavesdropper to get a copy.

This has been a long process. Facebook desperately wanted to convince me to correct the record, while at the same time not wanting to write something on their own letterhead (just a couple of comments, so far). I spoke at length with Privacy Policy Manager Nate Cardozo, whom Facebook hired last December from EFF. (Back then, I remember thinking of him — and the two other new privacy hires — as basically human warrant canaries. If they ever leave Facebook under non-obvious circumstances, we know that things are bad.) He basically leveraged his historical reputation to assure me that WhatsApp, and Facebook in general, would never do something like this. I am trusting him, while also reminding everyone that Facebook has broken so many privacy promises that they really can’t be trusted.

Final note: If they want to be trusted, Adam Shostack and I gave them a road map.

Hacker News thread.

EDITED TO ADD (8/4): Slashdot covered my retraction.