Online Differentially Private Conformal Prediction for Uncertainty Quantification

Traditional conformal prediction faces significant challenges with the rise of streaming data and increasing concerns over privacy. In this paper, we introduce a novel online differentially private conformal prediction framework, designed to construct dynamic, model-free private prediction sets. Unlike existing approaches that either disregard privacy or require full access to the entire dataset, our proposed method ensures individual privacy with a one-pass algorithm, ideal for real-time, privacy-preserving decision-making. Theoretically, we establish guarantees for long-run coverage at the nominal confidence level. Moreover, we extend our method to conformal quantile regression, which is fully adaptive to heteroscedasticity. We validate the effectiveness and applicability of the proposed method through comprehensive simulations and real-world studies on the ELEC2 and PAMAP2 datasets.

As more devices generate continuous streams of data—like smart meters tracking electricity or wearables monitoring heart rate—it becomes crucial to make accurate predictions while preserving personal privacy. Our research introduces a new method that enables real-time prediction without compromising individual data.Unlike traditional techniques that require storing and processing all past data, our approach uses each data point only once, making it highly efficient. To protect privacy, we add carefully designed randomness to each update, ensuring strong privacy guarantees while maintaining reliable predictive performance.This work offers a practical solution for real-time, privacy-preserving decision-making in sensitive domains such as healthcare, finance, and smart environments. It brings us one step closer to deploying safe and adaptive machine learning systems in everyday life.