Full Program

Machine Learning Model Complexity as a Mitigation Strategy Against Industrial Espionage through Membership Inference Attacks

Summary:

Machine learning (ML) models, particularly in the context of Federated Learning (FL), are increasingly used to enable predictive maintenance in smart industry. However, as these models become integral to operations, they also become potential targets for data leakage and membership inference attacks (MIAs). In this paper, we hypothesise that training on multi-dimensional data (i.e., multiple features instead of a single feature) improves resilience against MIAs compared to simpler single-feature models. To test this hypothesis we design an experimental testbed to empirically evaluate the vulnerability of ML models to black-box MIAs by training models of varying complexity on industrial time-series data. Additionally, we introduce a human expert's perspective to contextualise our findings in the realm of industrial espionage, highlighting the real-world implications of data leakage. Finally, we offer a set of observations and lessons learnt from controlled experiments discussing trade-offs between model complexity security and computational effort in industrial FL deployments.

Author(s):

Rustem Dautov    
SINTEF
Norway

Hui Song    
SINTEF
Norway

Christian Schaefer    
Ericsson AB
Sweden

Seonghyun Kim    
Ericsson AB
Sweden

Verena Pietsch    
FILL GmbH
Austria