Evaluating Privacy-Preserving Machine Learning in Critical Infrastructures: A Case Study on Time-Series Classification
This addresses privacy concerns in critical infrastructure applications like healthcare and energy, though it is incremental as it focuses on evaluating existing methods on underrepresented time-series data.
The study evaluated privacy-preserving machine learning methods for time-series data in critical infrastructures, finding that encryption is ineffective for deep learning, differential privacy is highly dataset-dependent, and federated methods are broadly applicable.
With the advent of machine learning in applications of critical infrastructure such as healthcare and energy, privacy is a growing concern in the minds of stakeholders. It is pivotal to ensure that neither the model nor the data can be used to extract sensitive information used by attackers against individuals or to harm whole societies through the exploitation of critical infrastructure. The applicability of machine learning in these domains is mostly limited due to a lack of trust regarding the transparency and the privacy constraints. Various safety-critical use cases (mostly relying on time-series data) are currently underrepresented in privacy-related considerations. By evaluating several privacy-preserving methods regarding their applicability on time-series data, we validated the inefficacy of encryption for deep learning, the strong dataset dependence of differential privacy, and the broad applicability of federated methods.