Byzantine-Robust Distributed Sparse Learning Revisited
For distributed machine learning systems, this work provides a robust and communication-efficient solution for high-dimensional sparse models under Byzantine failures.
The paper revisits Byzantine-robust distributed sparse learning, proposing a framework combining local ℓ1-regularized robust estimation with robust aggregation. It achieves near-optimal statistical rates and strong robustness under various Byzantine attacks, as confirmed by simulations.
We revisit Byzantine robust distributed estimation for high-dimensional sparse linear models. By combining local $\ell_1$-regularized robust estimation with robust aggregation at the server, the framework applies to pseudo-Huber regression, quantile regression, and sparse SVM. We show that the resulting estimators yield non-asymptotic guarantees and attain near-optimal statistical rates under mild conditions, while remaining communication-efficient. Simulations confirm strong robustness in estimation, support recovery and classification accuracy under various Byzantine attacks.