Unlearnable Examples for Time Series

Unlearnable examples (UEs) refer to training samples modified to beunlearnable to Deep Neural Networks (DNNs). These examples are usuallygenerated by adding error-minimizing noises that can fool a DNN model intobelieving that there is nothing (no error) to learn from the data. The conceptof UE has been proposed as a countermeasure against unauthorized dataexploitation 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, weintroduce the first UE generation method to protect time series data fromunauthorized training by deep learning models. To this end, we propose a newform of error-minimizing noise that can be selectively applied tospecific segments of time series, rendering them unlearnable to DNN modelswhile remaining imperceptible to human observers. Through extensive experimentson a wide range of time series datasets, we demonstrate that the proposed UEgeneration method is effective in both classification and generation tasks. Itcan protect time series data against unauthorized exploitation, whilepreserving their utility for legitimate usage, thereby contributing to thedevelopment of secure and trustworthy machine learning systems.