Search

Vertical federated learning (VFL) is attracting much attention because it enables cross-silo data cooperation in a privacy-preserving manner. While most research works in VFL focus on linear and tree models, deep models (e.g., neural networks) are not well studied in VFL. In this paper, we focus on SplitNN, a well-known neural network framework in VFL, and identify a trade-off between data security and model performance in SplitNN. Briefly, SplitNN trains the model by exchanging gradients and transformed data. On the one hand, SplitNN suffers from the loss of model performance since multiply parties jointly train the model using transformed data instead of raw data, and a large amount of low-level feature information is discarded. On the other hand, a naive solution of increasing the model performance through aggregating at lower layers in SplitNN (i.e., the data is less transformed and more low-level feature is preserved) makes raw data vulnerable to inference attacks. To mitigate the above trade-off, we propose a new neural network protocol in VFL called Security Forward Aggregation (SFA). It changes the way of aggregating the transformed data and adopts removable masks to protect the raw data. Experiment results show that networks with SFA achieve both data security and high model performance., Comment: Accepted by FL-IJCAI (https://federated-learning.org/fl-ijcai-2022/)

Botulinum neurotoxins are multifaceted molecules, which are truly unique not only in their mode of action, but also in their utility as a drug carrier either across the gut wall or to the nerve terminals. The molecule is divided into clear functional domains that can operate independently. This feature can be used to employ them as cargo carrier by linking other drugs or vaccines with the binding and translocation domains of BoNT. While the domain structures are largely independent of each other, the dynamic structure of these domains, especially that of the enzymatic domain (L chain), is quite different from the reported crystal structures for several BoNT serotypes and their enzymatic domain. This presentation will discuss the comparative structures of BoNT in crystal and solution for their relevance to the molecular mechanism of BoNT action. We have recently discovered that the enzymatically active structure of the BoNT exists as a molten-globule and that of the endopeptidase domain as a novel PRIME conformation, which may be relevant to the mechanistics of BoNT endopeptidase activity against SNAP-25 and to the development of inhibitors against it. This work was supported by DTRA under a DoD/Army Contract No. W911NF-06-1-0095.