Ensembler: Protect Collaborative Inference Privacy from Model Inversion Attack via Selective Ensemble
It addresses privacy risks for clients in cloud-based collaborative inference, offering a novel defense against adversarial attacks.
The paper tackles the problem of protecting sensitive client information from model inversion attacks during collaborative inference in cloud computing, introducing Ensembler, a framework that uses selective model ensemble to obfuscate reconstruction, resulting in up to 43.5% improvement in structural similarity with only 4.8% time overhead.
For collaborative inference through a cloud computing platform, it is sometimes essential for the client to shield its sensitive information from the cloud provider. In this paper, we introduce Ensembler, an extensible framework designed to substantially increase the difficulty of conducting model inversion attacks by adversarial parties. Ensembler leverages selective model ensemble on the adversarial server to obfuscate the reconstruction of the client's private information. Our experiments demonstrate that Ensembler can effectively shield input images from reconstruction attacks, even when the client only retains one layer of the network locally. Ensembler significantly outperforms baseline methods by up to 43.5% in structural similarity while only incurring 4.8% time overhead during inference.