PPDL: efficient dropout-resilient aggregation for privacy-preserving decentralized learning

Decentralized learning (DL) offers a novel paradigm in machine learning by distributing training across clients without central aggregation, enhancing scalability and efficiency. However, DL’s peer-to-peer model introduces unique challenges in protecting machine learning models against inference attacks and privacy leaks. By forgoing central bottlenecks, DL demands privacy-preserving aggregation methods to protect data from ‘honest but curious’ clients and adversaries, maintaining network-wide privacy. Privacy-preserving DL faces the additional hurdle of client dropout, clients not submitting updates due to connectivity problems or unavailability, further complicating the aggregation. In this paper, we propose three secret sharing-based dropout resilience protocols for privacy-preserving DL. We evaluate the efficiency, performance, and accuracy of these protocols through experiments on open datasets MNIST, Fashion-MNIST, SVHN, and CIFAR-10. We also compare our protocols with traditional secret-sharing solutions including a scenario that contains up to 1000 clients. Evaluations show that our protocols significantly outperform conventional methods, especially in scenarios with up to 30% of clients dropping out and model sizes of up to 10⁶ parameters. Our proposals demonstrate significantly high efficiency with larger models, higher dropout rates, and extensive client networks, highlighting their effectiveness in enhancing decentralized learning systems’ privacy and dropout robustness.