Google Dumps Interest-based Ad System For Another Interest-based Ad System

Google has given up on Federated Learning of Cohorts (FLoC), a categorization system for serving interest-based ads, and replaced it with Topics, a categorization system for serving interest-based ads.

Caught between the push to do something about cookie-based tracking and the counter-revolution to get regulators to keep third-party cookies alive, the Chocolate Factory has proposed a revision of its ill-fated FLoC plan.

"With Topics, your browser determines a handful of topics, like 'Fitness' or 'Travel & Transportation,' that represent your top interests for that week based on your browsing history," explained Vinay Goel, product director for Google's Privacy Sandbox, in a blog post on Tuesday.

In other words, your web history is broken down into hostnames, which are analyzed by a machine-learning algorithm to determine the general topics you're interested in. This all happens locally within your browser. Then websites aware of this tech can make an Google Topics API call to your browser to fetch an array of up to three topics derived from your visited sites that can then be used to present targeted ads.

As an ostensible nod to privacy, Topics are chosen on the user's device without the involvement of external servers – though ad networks and the like may be able to infer which topics prompted their ads – and are retained for only three weeks before being deleted. And Topics have been "thoughtfully curated to exclude sensitive categories, such as gender or race," we're told.

"When you visit a participating site, Topics picks just three topics, one topic from each of the past three weeks, to share with the site and its advertising partners," said Goel. "Topics enables browsers to give you meaningful transparency and control over this data, and in Chrome, we're building user controls that let you see the topics, remove any you don't like or disable the feature completely.

"Topics are selected entirely on your device without involving any external servers, including Google servers."

Yet even this latest gambit may come up short, at least as far as privacy advocates are concerned. Researchers have already established that a person's browsing history can be used to uniquely identify someone.

Introduced in 2020 and tested for a few months last year until privacy problems surfaced, FLoC was one of a number of proposals for preserving targeted advertising once third-party cookies get phased out for being irredeemable enablers of surveillance. 

Google refers to these proposals collectively as the Privacy Sandbox. They reflect not only the recognition that web browsers have real security and privacy gaps but also that tightening privacy laws in Europe and the US will not permit online advertising to continue in its current invasive form.

FLoC divides people into groups with similar interests, in order to hide individual identities in the crowd while still pitching people with marketing that corresponds to supposed interests. Google previously claimed that FLoC performs 95 per cent as well as cookie-based advertising.

However, a technical analysis of the proposal [PDF] published last June by Mozilla's Eric Rescorla and Martin Thomson suggests Google's technology fails to provide sufficient anonymity.

"In particular, it may be possible to identify individual users by using both FLoC IDs and relatively weak fingerprinting vectors," the Mozillans concluded. "When considered as coexisting with existing state-based tracking mechanisms, FLoC has the potential to significantly increase the power of cross-site tracking."

At the same time, Google has been set upon by ad industry challengers who argue that browser cookies should be left alone because their businesses rely on cookie-based data collection. Revanchist ad firms have been trying to convince antitrust regulators that removing third-party cookies will only make Google more powerful because Google will retain ways to gather data while others will not.

The changes are anticompetitive because they raise barriers to entry and exclude competition in the exchange and ad buying tool markets

That line of argument is reflected in the antitrust complaints Google is facing in the US, such the one filed by Texas and other states, which describes the Privacy Sandbox proposals thus: "Overall, the changes are anticompetitive because they raise barriers to entry and exclude competition in the exchange and ad buying tool markets, which will further expand the already dominant market power of Google's advertising businesses."

In a blog post responding to Google's announcement, Peter Snyder, senior director of privacy at competing browser maker Brave, characterized Topics as a rebranding of FLoC that fails to address important privacy issues.

Topics, he said, differs from FLoC by conditionally broadcasting interests rather than making them available on demand and by adding some randomness to make user reidentification via fingerprinting techniques a bit more challenging.

Even so, he argues these changes still amount to putting lipstick on a pig. Both FLoC and Topics, he contends, hurt privacy and competition. The choice, he argues, should not be the lesser of two evils.

"Google's proposals are privacy-improving only from the cynical, self-serving baseline of 'better than Google today," Snyder said. "Chrome is still the most privacy-harming popular browser on the market, and Google is trying to solve a problem they introduced by taking minor steps meant to consolidate their dominance on the ad tech landscape. Topics does not solve the core problem of Google broadcasting user data to sites, including potentially sensitive information." ®