Understanding and Controlling User Linkability in Decentralized Learning

14 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Cronus: Robust and Heterogeneous Collaborative Learning with Black-Box Knowledge Transfer (2019, background): “Although federated learning prevents its participants from sharing their private data, recent works [40], [42], [36] demonstrate passive and active inference attacks against this setting, and successfully infer sensitive information about the parties’ private data.”
MLCapsule: Guarded Offline Deployment of Machine Learning as a Service (2018, background): “Local execution also avoids the Internet communication between the user and the MLaaS provider and eliminates possible attack surfaces due to network communication [20, 24].”
DeepChain: Auditable and Privacy-Preserving Deep Learning with Blockchain-Based Incentive (2019, background): “al [20] exploited the intermediate gradients to launch linkability attack on training data, since the gradients contain sufficient data features.”
Preserving Data Privacy via Federated Learning: Challenges and Solutions (2020, methodology): “To mitigate the risks of linkability attacks, according to various strategies of Orekondy et al. [8], it is required to reduce the distinctiveness in model updates by using calibrated domain-speciﬁc data augmentation.”