# Federated Learning of Cohorts > type of web tracking based on browsing history **Wikidata**: [Q106522497](https://www.wikidata.org/wiki/Q106522497) **Wikipedia**: [English](https://en.wikipedia.org/wiki/Federated_Learning_of_Cohorts) **Source**: https://4ort.xyz/entity/federated-learning-of-cohorts ## Summary Federated Learning of Cohorts (FLoC) is a type of web tracking software and client-side web API designed to group users based on their browsing history for the purpose of targeted advertising. It utilizes SimHash algorithms to classify users into cohorts, operating as a privacy-focused alternative to traditional cookie-based tracking. ## Key Facts - **Acronym:** FLoC - **Classification:** Instance of web tracking, software, and client-side web API. - **Primary Use:** Targeted advertising. - **Algorithm:** Uses SimHash for clustering. - **Source Code:** Hosted on GitHub at `https://github.com/WICG/floc`. - **License:** Dual-licensed under the W3C Software License and the 3-clause BSD License. - **Chrome Status:** Feature ID `5710139774468096` on the Chrome platform status tracker. - **Wikipedia Presence:** Available in 5 languages (German, English, French, Japanese, Russian). ## FAQs ### Q: What is the primary purpose of Federated Learning of Cohorts? A: The primary purpose of FLoC is to facilitate targeted advertising by grouping users with similar browsing histories into "cohorts," thereby allowing advertisers to target groups rather than individuals. ### Q: Is FLoC considered a software or a tracking method? A: It is both. Structured data classifies FLoC as "software" (a non-tangible executable component) and specifically as a "client-side web API" that functions as a method of "web tracking." ### Q: What technology does FLoC use to group users? A: FLoC uses SimHash, a hashing algorithm, to analyze browsing history and assign users to specific cohorts. ## Why It Matters Federated Learning of Cohorts represents a significant technical shift in the digital advertising ecosystem, specifically within the Google Chrome infrastructure. As a client-side web API, it attempts to reconcile the commercial necessity of targeted advertising with growing privacy concerns by processing data locally on the user's device rather than sharing individual browsing history with third parties. Its development has sparked considerable debate within the tech community, evidenced by references such as the Electronic Frontier Foundation's critique titled "Google’s FLoC Is a Terrible Idea" and Wired magazine's coverage of rivals fighting back against Chrome's cookie plans. By moving tracking logic to the browser itself, FLoC challenges the established "status quo" of web tracking and serves as a focal point for discussions regarding the future of user privacy and the economics of the open web. ## Notable For - **Hybrid Classification:** It is uniquely classified as both a web tracking practice and a specific software component/API. - **SimHash Implementation:** It is notable for its specific application of SimHash technology for interest-based clustering. - **Open Source Licensing:** The project is notable for being available under dual open-source licenses (W3C and BSD). - **Controversy:** It is distinguished by immediate, high-profile criticism from privacy advocates, such as the EFF, upon its proposal. ## Body ### Technical Definition and Classification Federated Learning of Cohorts (FLoC) is defined as a mechanism for web tracking based on a user's browsing history. Technically, it functions as a client-side web API and a piece of software. Unlike traditional server-side tracking, FLoC calculates a user's interests directly on their device (client-side) without exporting the raw browsing data to a server. ### Functionality and Algorithm The system operates by grouping users into "cohorts"—groups of thousands of users who share similar browsing habits. This grouping is facilitated using **SimHash**, a technique referenced in the source data. The output of this process is utilized specifically for **targeted advertising**, allowing advertisers to serve ads to specific interest groups without needing to know the specific identity or individual browsing history of a single user. ### Development and Availability FLoC was developed within the Web Incubator Community Group (WICG) context. The source code repository is hosted on GitHub (`https://github.com/WICG/floc`) and is made available under the **W3C Software License** and the **3-clause BSD License**. The official technical documentation and proposal are accessible via the WICG website (`https://wicg.github.io/floc/`). ### Context and Criticism The introduction of FLoC has been met with scrutiny regarding its privacy implications. The source data indicates significant pushback, citing references such as the article "Google’s rivals are fighting back against Chrome’s big cookie plan" and the EFF's publication "Google’s FLoC Is a Terrible Idea." Despite its classification as a tool to enhance privacy compared to third-party cookies, critics argue it may introduce new fingerprinting vectors or unsanctioned tracking avenues. The feature is tracked in the Chrome Platform Status under Feature ID `5710139774468096`. ## References 1. [Source](https://github.com/WICG/floc/blob/main/LICENSE.md) 2. [Google’s rivals are fighting back against Chrome’s big cookie plan. 2021](https://www.wired.co.uk/article/google-chrome-cookie-alternatives) 3. [What is FLoC?. 2021](https://web.dev/floc/) 4. [Google’s FLoC Is a Terrible Idea. Electronic Frontier Foundation. 2021](https://www.eff.org/deeplinks/2021/03/googles-floc-terrible-idea)