Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unlearnable Examples For Time Series

Yujing Jiang, Xingjun Ma, Sarah Monazam Erfani, James Bailey

TL;DR

The paper addresses protecting time-series data from unauthorized deep learning by extending Unlearnable Examples to the temporal domain. It introduces a selective, error-minimizing noise framework that targets crucial time-series segments via a min-min bilevel optimization and a controllable region vector, preserving non-sensitive data. Empirical results show strong protection for both time-series classification and generation tasks, with large accuracy drops using noise on small data fractions and substantial degradation of generative models via TSTR. This approach offers a practical privacy-preserving mechanism for sharing time-series data without compromising legitimate use cases, contributing to secure and trustworthy ML systems.

Abstract

Unlearnable examples (UEs) refer to training samples modified to be unlearnable to Deep Neural Networks (DNNs). These examples are usually generated by adding error-minimizing noises that can fool a DNN model into believing that there is nothing (no error) to learn from the data. The concept of UE has been proposed as a countermeasure against unauthorized data exploitation on personal data. While UE has been extensively studied on images, it is unclear how to craft effective UEs for time series data. In this work, we introduce the first UE generation method to protect time series data from unauthorized training by deep learning models. To this end, we propose a new form of error-minimizing noise that can be \emph{selectively} applied to specific segments of time series, rendering them unlearnable to DNN models while remaining imperceptible to human observers. Through extensive experiments on a wide range of time series datasets, we demonstrate that the proposed UE generation method is effective in both classification and generation tasks. It can protect time series data against unauthorized exploitation, while preserving their utility for legitimate usage, thereby contributing to the development of secure and trustworthy machine learning systems.

Unlearnable Examples For Time Series

TL;DR

The paper addresses protecting time-series data from unauthorized deep learning by extending Unlearnable Examples to the temporal domain. It introduces a selective, error-minimizing noise framework that targets crucial time-series segments via a min-min bilevel optimization and a controllable region vector, preserving non-sensitive data. Empirical results show strong protection for both time-series classification and generation tasks, with large accuracy drops using noise on small data fractions and substantial degradation of generative models via TSTR. This approach offers a practical privacy-preserving mechanism for sharing time-series data without compromising legitimate use cases, contributing to secure and trustworthy ML systems.

Abstract

Unlearnable examples (UEs) refer to training samples modified to be unlearnable to Deep Neural Networks (DNNs). These examples are usually generated by adding error-minimizing noises that can fool a DNN model into believing that there is nothing (no error) to learn from the data. The concept of UE has been proposed as a countermeasure against unauthorized data exploitation on personal data. While UE has been extensively studied on images, it is unclear how to craft effective UEs for time series data. In this work, we introduce the first UE generation method to protect time series data from unauthorized training by deep learning models. To this end, we propose a new form of error-minimizing noise that can be \emph{selectively} applied to specific segments of time series, rendering them unlearnable to DNN models while remaining imperceptible to human observers. Through extensive experiments on a wide range of time series datasets, we demonstrate that the proposed UE generation method is effective in both classification and generation tasks. It can protect time series data against unauthorized exploitation, while preserving their utility for legitimate usage, thereby contributing to the development of secure and trustworthy machine learning systems.
Paper Structure (17 sections, 5 equations, 1 figure, 2 tables)

This paper contains 17 sections, 5 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data Poisoning
  4. Adversarial Attack
  5. Unlearnable Examples
  6. Error-minimizing Noise For Time Series
  7. Objective
  8. Threat Model
  9. Challenges
  10. Problem Formulation
  11. A Straightforward Baseline Approach
  12. Controllable Noise on Partial Time Series Samples
  13. Experiments
  14. Experiment Setup
  15. Against Classification Models
  16. ...and 2 more sections

Figures (1)

  • Figure 1: Illustration of the control vector applied on a time series sample of length $T$. Data protection is indicated when the control vector highlights particular time stamps with a value of 1 (marked in black).