English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Gradient inversion attacks on federated learning systems reconstruct clienttraining data from exchanged gradient information. To defend against suchattacks, a variety of defense mechanisms were proposed. However, they usuallylead to an unacceptable trade-off between privacy and model utility. Recentobservations suggest that dropout could mitigate gradient leakage and improvemodel utility if added to neural networks. Unfortunately, this phenomenon hasnot been systematically researched yet. In this work, we thoroughly analyze theeffect of dropout on iterative gradient inversion attacks. We find that stateof the art attacks are not able to reconstruct the client data due to thestochasticity induced by dropout during model training. Nonetheless, we arguethat dropout does not offer reliable protection if the dropout inducedstochasticity is adequately modeled during attack optimization. Consequently,we propose a novel Dropout Inversion Attack (DIA) that jointly optimizes forclient data and dropout masks to approximate the stochastic client model. Weconduct an extensive systematic evaluation of our attack on four seminal modelarchitectures and three image classification datasets of increasing complexity.We find that our proposed attack bypasses the protection seemingly induced bydropout and reconstructs client data with high fidelity. Our work demonstratesthat privacy inducing changes to model architectures alone cannot be assumed toreliably protect from gradient leakage and therefore should be combined withcomplementary defense mechanisms.

Dropout is NOT All You Need to Prevent Gradient Leakage