Continual Learning with Differential Privacy

2021/10/11 by Pradnya Desai, Phung Lai, Desai, Pradnya +5 · 3 citations
Computer Science · #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.CR #cs.LG

paper · pdf · doi:10.48550/arxiv.2110.05223

The paper will appear at ICONIP21

arxiv created 2021/10/11 · arxiv updated 2021/10/12

Abstract

In this paper, we focus on preserving differential privacy (DP) in continual learning (CL), in which we train ML models to learn a sequence of new tasks while memorizing previous tasks. We first introduce a notion of continual adjacent databases to bound the sensitivity of any data record participating in the training process of CL. Based upon that, we develop a new DP-preserving algorithm for CL with a data sampling strategy to quantify the privacy risk of training data in the well-known Averaged Gradient Episodic Memory (A-GEM) approach by applying a moments accountant. Our algorithm provides formal guarantees of privacy for data records across tasks in CL. Preliminary theoretical analysis and evaluations show that our mechanism tightens the privacy loss while maintaining a promising model utility.

Cited by

• Forget What's Sensitive, Remember What Matters: Token-Level Differential Privacy in Memory Sculpting for Continual Learning
• What is the role of memorization in Continual Learning?
• Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay

Related

• Privacy Leakage via Output Label Space and Differentially Private Continual Learning
• Forget What's Sensitive, Remember What Matters: Token-Level Differential Privacy in Memory Sculpting for Continual Learning
• Lifelong DP: Consistently Bounded Differential Privacy in Lifelong Machine Learning
• Three Tools for Practical Differential Privacy
• A General Approach to Adding Differential Privacy to Iterative Training Procedures