Secure aggregation consists of computing the sum of data collected from multiple sources without disclosing these individual inputs. Secure aggregation has been found useful for various applications ranging from electronic voting to smart grid measurements. Recently, federated learning emerged as a new collaborative machine learning technology to train machine learning models. In this work, we study the suitability of secure aggregation based on cryptographic schemes to federated learning. We first provide a formal definition of the problem and suggest a systematic categorization of existing solutions. We further investigate the specific challenges raised by federated learning and analyze the recent dedicated secure aggregation solutions based on cryptographic schemes. We finally share some takeaway messages that would help a secure design of federated learning and identify open research directions in this topic. Based on the takeaway messages, we propose an improved definition of secure aggregation that better fits federated learning.

In this paper, we propose to study privacy concerns raised by the analysis of Electro CardioGram (ECG) data for arrhythmia classification. We propose a solution named PAC that combines the use of Neural Networks (NN) with secure two-party computation in order to enable an efficient NN prediction of arrhythmia without discovering the actual ECG data. To achieve a good trade-off between privacy, accuracy, and efficiency, we first build a dedicated NN model which consists of two fully connected layers and one activation layer as a square function. The solution is implemented with the ABY framework. PAC also supports classifications in batches. Experimental results show an accuracy of 96.34% which outperforms existing solutions.

No abstract