Comment by cyphers | Hacker News Reader

Comment by cyphers | Hacker News Reader

Parent Full thread

The passage quoted above refers to the way google plans to audit FLoC, not FLoC itself.

Each browser is assigned a single floc ID, which is (for now) a 50-bit hash that is not human readable on its own. But the floc ID is intended to capture something meaningful about browsing behavior. Two users with similar browsing history are likely to have similar, if not the same, floc IDs. Floc IDs which are close together (hamming distance-wise) also reflect similar browsing histories. (http://matpalm.com/resemblance/simhash/)

It will be totally possible to gather data about what kind of people belong to a particular floc -- provided that you operate a large website or ad network. It will probably not be possible for normal people.

There's a risk that advertisers will learn that a particular floc ID corresponds to a "sensitive category" of users. That's what google is trying to prevent.

Google plans to audit floc IDs for correlation with visits to sensitive websites. In order to do so, it's assigning a "sensitivity" label to certain websites, and then applying that label to each user who visits those websites. Then it will run analyses to see whether users with a particular floc ID are more likely to have a particular label. Floc IDs that correlate too strongly with a particluar label will be thrown out.

The passage is attempting to explain why this does not prevent adtech from learning other meaningful things about flocs -- for example, that users in a specific floc are disproportionately female, or Muslim, or suffering from depression.

As for whether FLoC is in chrome, see the origin trial page here: https://developer.chrome.com/origintrials/#/view_trial/21392...

All good points, but they don't change the conclusion I wrote though, you effectively further specified the lengths to which Google is going to prevent any potential bad correlation to sensitive categories. Of course there will be correlation of some sort, they aren't random groups, but so long as you can't tell sensitive categories, you can only observe that a group seems more interested in a particular advertiser products. You can't tell their browsing history, just their interests.

Regarding the trial, yeah I know that origin thing is up, and the minutes for yesterday's W3C IWABG meeting aren't going to be out until next week, but the team working on this stuff at Google said that the trial up there is kind of fake and the code isn't merged in Chrome yet and won't be until 91, at some point I have to trust the engineers working on this stuff more than the rest.

[edit]: To be clear, I don't mean that there's no experiment going on, I just mean that it's not the browser running it, that the removal of 3rd party cookies as a way to opt-out likely is due to the fact that they may be relying on server-side computation to do this experiment rather than the browser, I wouldn't see why the browser would care about cookies to track your history, it's already stored in the browser.

> Of course there will be correlation

Yes, and given enough data this:

> so long as you can't tell sensitive categories

will not be true.

> I have to trust the engineers

The engineers that work for a giant corporation whose entire reason for existing is mining and collating as much data as possible on you? Why would you trust them at all?

Because they aren't lying... The source code of this stuff is available in the open and you can verify it. There's no FLoC in the merged code in the main branch.

The point, I think, isn't that this is currently in Chrome, but rather that there are plans to add it to Chrome in the future.

Even with Google's auditing, there are sensitive things that Google (hopefully) can't detect, but others can.

For example, insurance claims. I hope Google isn't tracking enough information to test for correlations between FLoC cohorts and insurance claims, but insurers will have access to that info when customers visit their site to pay bills.

Perhaps they'll find they can more accurately anticipate customer's claims from their FLoC cohort, and use that to price new policies, thus indirectly pricing policies based on browsing history.

The FLoC cohort changes pretty frequently with your history, I don't think you will be tagged for life with "insurance claims", but you could, for a period of time, show up as part of a group that looks interested in insurance, it will hardly go much deeper than this.

To offer just one of many possible examples that come to mind: the FLoC cohort might cluster people who read articles about becoming parents, which, though not visible in the data Google has, correlates with future claims for fertility treatments, pregnancy, etc.

The insurance company has the data, and it doesn't even need to know any of the details, it just needs to realize that certain cluster numbers correlate with future claims.

Perhaps Google will think of some of these correlations, and flag articles about becoming parents as sensitive, but how many unexpected correlations might be overlooked?

The only way to discover this is to buy traffic from that cohort to actually test the behavior. That being said the first version of floc uses just the domain name of your history.

I’m not sure how the insurance company gets that data though in your example.

> I’m not sure how the insurance company gets that data though in your example.

AIUI, they can get customers' FLoC cohort when customers visit their site to pay bills (or do anything else), and see prospective customers' FLoC cohort when issuing quotes online.

Who knows what they'll be able to learn from that? I certainly don't, but I expect them to try, and that worries me.

Reply on news.ycombinator.com