SASH: Efficient Secure Aggregation Based on SHPRG For Federated Learning
This addresses privacy protection in Federated Learning systems, particularly for scenarios with many clients and large models, though it appears to be an incremental improvement on existing secure aggregation methods.
The paper tackles the problem of private training data leakage in Federated Learning systems by proposing SASH, a secure aggregation scheme based on seed homomorphic pseudo-random generators. The result is a 20x efficiency improvement over baseline while preserving privacy against worst-case colluding adversaries and handling client dropouts without extra overhead.
To prevent private training data leakage in Fed?erated Learning systems, we propose a novel se?cure aggregation scheme based on seed homomor?phic pseudo-random generator (SHPRG), named SASH. SASH leverages the homomorphic property of SHPRG to simplify the masking and demask?ing scheme, which for each of the clients and for the server, entails an overhead linear w.r.t model size and constant w.r.t number of clients. We prove that even against worst-case colluding adversaries, SASH preserves training data privacy, while being resilient to dropouts without extra overhead. We experimentally demonstrate SASH significantly improves the efficiency to 20 times over baseline, especially in the more realistic case where the numbers of clients and model size become large, and a cer?tain percentage of clients drop out from the system.