Google taps Fastly to make cookie-free adtech FLEDGE fly
- Reference: 1678951810
- News link: https://www.theregister.co.uk/2023/03/16/google_fledge_privacy_sandbox/
- Source link:
Google's [1]FLEDGE is a Privacy Sandbox proposal to allow [2]remarketing and custom audiences. It aims to let websites present ads that reflect visitor interests, without allowing visitors to be tracked or identified.
FLEDGE works like this: When a user visits a website associated with an interest group, like hiking, an interest group owner (a demand side platform, or DSP, enabling the purchase of ads associated with the site) can request the user's browser to join the relevant interest group, with a limited lifespan, using a JavaScript function.
[3]
If the JavaScript call succeeds – it could fail, be blocked, or be refused – the browser stores the interest group name (for example, hiking), the URL of the interest group owner, and configuration data to allow the browser to participate in an auction to place an ad.
[4]
[5]
When the user visits another website selling ads, the seller of that ad space – typically a supply side platform, or SSP – has the option to use FLEDGE to run an ad auction for an interest-relevant ad.
The SSP makes another JavaScript call to begin the auction within the browser – that is, the code fetches a list of interest group owners stored in the user's browser and invites some of them to bid to show the user a targeted ad.
[6]
Bidding is done via the bidding logic URL specified in the configuration data, which gets supplied with the interest group and information about the ad seller (the site's SSP or the site itself). The seller then receives the bids and displays the winning ad in a [7]fenced frame – the (hopefully) secure successor to the iframe.
FLEDGE is being [8]tested right now in Google's Chrome browser. Other browser vendors have yet to declare whether or not they will implement Google's Privacy Sandbox APIs.
Sort-of anonymity online? We'll see
To make this work while ensuring privacy, Google is running servers that implement a technique called [9]k -anonymity . It's a way to promote privacy by hiding individuals within a crowd, the size of which is represented by the variable k .
FLEDGE applies k -anonymity [10]to several aspects of the bid process. For example, an ad provider (DSP) could create a unique and thus trackable FLEDGE group (such as hikingUser23).
To prevent that, FLEDGE won't let a browser set an interest group unless there are at least k other browsers trying to set that same interest group. And to prevent ads targeting individuals, FLEDGE applies k -anonymity to ad rendering URLs, so a crowd of at least 50 users per ad design, within the past seven days, is required for an ad to be shown.
[11]
To make this work in a way that hides potentially identifying information – like a website visitor's IP address and the browser's User-Agent string – Google is putting its k -anonymity servers behind a third party. This is where Fastly comes in, running an [12]Oblivious HTTP (OHTTP) relay .
[13]Google ready to kick the cookie habit by Q3 2024, for real this time
[14]Google lets a few Android devices into its Privacy Sandbox
[15]Shot down: Google's grand fancy plan for pro-privacy targeted ads
[16]Google postpones Chrome's third-party cookie bonfire yet again
As Google software engineer Philip Lee explains in a [17]blog post , the user's Chrome browser sends an encrypted request through the OHTTP relay to Google's k -anonymity servers.
"The relay therefore doesn't see the content of the request but is aware of the user's IP address," Lee explains. "Conversely, the k -anonymity server (and gateway) are unaware of the user's identity but can see the content of the request."
There are some limitations to the privacy afforded by this approach. One is that the website publisher can still see the IP addresses of visitors. The other is that Google will have lots of identifying information about an individual if that person is signed into a Google Account through Chrome. But Google at least insists it will apply its k -anonymity protection to its own advertising services.
This approach "will offer better privacy since source address IP would be masked," said Lukasz Olejnik, independent privacy researcher and consultant, in an email to The Register .
"I know from my research that IP addresses, along with other information, [are] a strong identifier. It will certainly be a bit more difficult to do privacy research in a few years from now. That said, I'd like to think that my [18]previous works somewhat contribute to the current evolution, which is positive!"
Olejnik expects Google's approach will lead to a more formalized web ad infrastructure platform. "The assumption here is that the platform would remain open to all competitors," he said. "If I understand correctly, this is the core premise behind the UK [Competition and Markets Authority] process."
Asked about whether Fastly might be in a position to abuse its role as a trusted intermediary, Olejnik replied, "Fastly as the management of the partial infrastructure should be trusted in this case. That said, I'm generally cautious about such centralized systems. They should always be done with care. We will only be able to review a final proposal and a design, once they will come."
As for the security of Oblivious HTTP, Olejnik allowed room for skepticism but said it's a [19]technical standard and that it's used, with Fastly, to support Apple's Private Relay.
"The key question is whether any other infrastructural providers would want to join in, too," he said. ®
Get our [20]Tech Resources
[1] https://developer.chrome.com/docs/privacy-sandbox/fledge/
[2] https://developer.chrome.com/docs/privacy-sandbox/glossary/#remarketing
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZBL2zWbNPF8Pz@gi7T8YYAAAAM0&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZBL2zWbNPF8Pz@gi7T8YYAAAAM0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZBL2zWbNPF8Pz@gi7T8YYAAAAM0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZBL2zWbNPF8Pz@gi7T8YYAAAAM0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://developer.chrome.com/docs/privacy-sandbox/fenced-frame/
[8] https://chromestatus.com/feature/5733583115255808
[9] https://github.com/WICG/turtledove/blob/main/FLEDGE_k_anonymity_server.md
[10] https://github.com/WICG/turtledove/blob/main/FLEDGE_k_anonymity_server.md#use-cases-in-the-privacy-sandbox
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/personaltech&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZBL2zWbNPF8Pz@gi7T8YYAAAAM0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[12] https://github.com/WICG/turtledove/blob/main/FLEDGE_k_anonymity_server.md#oblivious-http
[13] https://www.theregister.com/2023/02/01/google_cookie_sandbox/
[14] https://www.theregister.com/2023/02/14/google_privacy_sandox_beta/
[15] https://www.theregister.com/2023/01/18/google_topics_api/
[16] https://www.theregister.com/2022/07/28/google_postpones_thirdparty_cookie_phaseout/
[17] https://developer.chrome.com/blog/oblivious-http-for-k-anon-server-with-fastly/
[18] https://lukaszolejnik.com/works
[19] https://www.ietf.org/archive/id/draft-thomson-http-oblivious-01.html
[20] https://whitepapers.theregister.com/
For GDPR they aren't allowed to reduce what you see if you don't consent... which seems to me a good thing. The reasoning being that it isn't a real choice otherwise.
As for this system, why would anyone click 'please track me' and again in GDPR land I would think the default setting would have to be "don't track".... which incidently we've already got a setting for...
GDPR has sometimes been a sledgehammer to crack a nut. That's not to say something needs/needed to be done about privacy but, like a lot of targeted legislation, it hurts those who try to act responsibly more than those who are happy bending/breaking rules.
I would have no objection to content being restricted if I reject tracking and adverts - nothing comes free and if advertising revenue is needed to maintain the site, then so be it. It need be no different to sites that require a subscription for full access. My local paper (that I can download free via my library account) has an online version where some articles are open and others require a log-in (after a trial period). If the site has value, then people need to be prepared to accept the cost - in this case, being served adverts than paying a subscription.
Just keep it simple
Ads related to the content of the website I am on are the best option for sites that want to serve ads.
The "interest" concept is too flawed.
If I am briefly visiting a lot of sites looking at washing machines (e.g. I need to replace a broken one that is beyond repair) and then stop visiting washing machine related sites (because I have purchased one) then, even though "interests" are supposed to be short lasting it would mean that for a while I potentially still get served ads for washing machines, no use to me, or the advertiser who has wasted their cash on those ads.
Similarly for lots of other things I may be needing to purchase but have zero need for ads relating to them once my purchase is over (apart from a relatively small number of sites I visit regularly, most of my web browsing is things to purchase & often not even an "interest" of mine as often partner has tasked me to order something). Live "in the sticks" and so in many cases online orders for more "specialist" items not available locally (such as a washing machine) less hassle to do that than a 50 mile round trip to nearest big city to look around "white goods warehouses" or whatever item I am looking to buy.
Am I reading this right?
The auction runs in my browser, using my CPU, internet and electricity to provide a service I am neither interested in or want?
Or does the browser just trigger the process?
Re: Am I reading this right?
That's what I was thinking. But then it occurred to me, that's the way it's always been -- we pay for our own oppression.
Let's create the "I'm not interested in ads and don't want to be tracked" interest group.
Is it just me, or has the internet been ruined by the advertising industry? Sure, it helps to "pay" for "free" sites, but it seems to have become the only purpose for a lot of sites.
Re: Let's create the "I'm not interested in ads and don't want to be tracked" interest group.
I see the unfiltered web on my non-techie friends' phones and it is the substance of nightmares. Cookie 'consent'/Popups/Autoplaying videos/Ads strewn everywhere. Javascript SPAs for news websites. Everything 'appified' to enable even easier tracking through $PHONE_OS APIs rather than in a plain old browser.
We can't put the clock back, but we can pray for its collapse under its own weight and ensure we're at a safe distance when it finally caves in.
noscript ftw.
Won't be fooled again
> targeted ads
> privacy
Pick one.
If I don't want to be tracked, it should be as simple as a default off switch in the browser and then exclusions I make, not by a company. If they track unlawfully, then massive fines based on global revenue, if I don't allow them to track, then they have the right to reduce the content that I can see but I also have the right to go elsewhere. Eventually, a balance will naturally occur.