Homepage
Welcome to CLiP Protocol
The CLiP protocol (context-aware Local Information Protection Protocol) is a novel solution for frequency estimation with personalized privacy budgets.
It empowers individual data owners — such as students — to protect their personal data while still enabling meaningful learning analytics.
Unlike traditional Local Differential Privacy (LDP) approaches, which assign a uniform privacy budget to all users, the CLiP Protocol introduces a personalized privacy-by-design approach. This ensures that individual characteristics and privacy preferences are respected without sacrificing the utility of the collected data.
✨ Why CLiP Protocol?
In conventional LDP implementations, data collectors define a general privacy budget for all participants, regardless of personal sensitivity or data variability. This "one-size-fits-all" model is insufficient when users have diverse privacy needs.
CLiP Protocol solves this by: - Allowing each individual to select their privacy level. - Preserving critical data utility for learning analytics. - Supporting privacy without relying on trusted servers.
🛠️ How It Works
The CLiP Protocol relies on privacy sketching and LDP techniques for sequential data events like: - Student activity logs - Click counts - Interaction events
The process is divided into two main components:
| Client Side | Server Side | |
|---|---|---|
| Runs on the data owner's device | Runs on the data collector’s server | |
| Preprocesses and privatizes data | Aggregates privatized data and handles queries | |
| Applies the privacy mechanism locally | Updates frequency sketches and responds to frequency queries |
The server is treated as untrusted: it only receives already privatized data.
👥 Main Actors
-
Data Owner: An individual providing raw information (e.g., students, researchers).
-
Data Collector: Institutions like universities managing the privatized data.
-
Data Consumer: External researchers or parties requesting frequency-based insights.
🧩 Workflow Stages
The CLiP Protocol operates through four functional stages:
-
Setup: Configuration of privacy parameters prior to data collection.
-
Mask: Client-side anonymization of raw data based on personalized privacy budgets.
-
Aggregation: Server-side update of frequency sketches with privatized input.
-
Estimation: Server-side frequency estimation and response to queries.
Only the Mask stage occurs on the client side, ensuring that raw data never leaves the owner’s device unprotected.
The figure above illustrates the complete CLiP Protocol workflow, showing how privacy and utility are balanced across client and server components.