Hacker News Reader

Comment by photochemsyn | Hacker News Reader

It's a relatively invasive technique that requires a computer's camera to record details of a person's face, and that's marketable data that any commercial shopping site would likely collect and sell on to other parties. It'd be much simpler to implement arrow key / spacebar navigation through a 3D scene (still computationally expensive relative to rendering the standard HTML/CSS/JS page), but then that data couldn't be collected.

In general interactive 3D on the web still uses too much computational power on the client-side for most devices.

View on HN

It is not imaging your face like a normal camera would. It is using an infrared camera to pick out the location of two dots (the relative position and distance of the two dots in the camera”s field of view). That information stream could be compressed to a few bytes per second without losing tracking fidelity. That does not seem very invasive. Even if a high resolution pixel stream were to be sent from the camera to an untrustworthy, persistence layer, it would be infrared, which doesn’t seem very privacy-invasive.

> invasive technique

It wouldn't be so bad if it were a browser feature, i.e. a website can use an "Eye Tracking" API instead of the Camera API (with permission granted), and get limited data.

But of course, just like gait detection and etc... there would be a whole genre of preprints dedicated to eye tracking fingerprinting

Couldn’t this sort of thing be built into the browser which could then just expose an API for head orientation and eye position without any privacy invading details?

Physically you'd still need to have the camera on, and having the "your webcam is in use" LED become ever-presently on defeats the ability to detect when the camera is actually being used as a camera.

"Details of a person's face is marketable data".

Okay, it's not that I don't believe you, but what do they do with that data? How does it help? Unless you're marketing acne cream, I suppose.

People who fit into <some racial/ethnic/age/gender bucket> spend more time looking at <x products>, fed to advertisers who can tie it to swathes of your other online activity

Even if it’s _theoretically_ coarse data/bucketing, that adds up to a lot of data to help deanonymization and targeting. Even if that’s done in the spirit of “just for ads”, it’s not a big leap to political targeting etc, and with the right data leaks, potentially even actual harassment.

Uniquely and reliably track a single individual’s web browsing habits.

Is that not the holy grail of adtech?

Some bad actor would check out what is in your apartment and then market to you.

Say they see you have legos, they market legos.

Or they see what you have (say a gun) so they rob you.

Or they track your habits. We noticed your apartment is dirty/messy - pay us a ransom or we post on linkedin or call CPS. Or use our cleaning service.

Also I bet you have to walk fully dressed when browsing (especially women) and you need to shave, or your social score drops.

Ah and will you know if the website ever stopped watching? Is it watching you (and recording) having sex?

Knowing exactly where the user is looking is pretty good data I would say.

"User looked at the buy button 5 times in the session but they did not buy" and so on.

An awful lot of us already wear glasses, whether prescription or blue light blocking.

If tiny IR lights or something else distinctive that the camera could recognize on the glasses could be used instead of actual eye/face tracking, I'll bet even more people would wear them.

Imagine the possibilities! If you have enough facial expressions, looked for n ms, moved head n yards, with enough a-b testing you could dynamically design products. Like Weird Science only worse.

Reply on news.ycombinator.com