Survey maps security risks in distributed and federated learning
Original title: Trustworthy Distributed Machine Learning in Cloud-Edge-IoT Systems: A Heterogeneity-Aware Survey of Security, Privacy, and Robustness
Federated learning is treated as a high-risk case, with defenses compared from aggregation to trusted execution.
A survey proposes a heterogeneity-aware roadmap for trustworthy distributed machine learning across cloud-edge-IoT systems. It examines assets, adversaries, trust assumptions, and how non-IID data, resource asymmetry, and intermittent participation affect attacks and defenses. Federated learning is presented as a prominent high-risk case. The authors compare layered defenses and identify evaluation and governance gaps.
- Reviewed attacks include poisoning, backdoors, Byzantine manipulation, and model extraction.
- Defenses include robust aggregation, differential privacy, secure aggregation, and trusted execution environments.
- Open issues include non-IID benchmarks, privacy budgets, and client-level disparity.
From the original article
1 min read at arxiv.orgDistributed machine learning (DML), including conventional distributed training and federated learning (FL), enables collaborative model development across cloud-edge-IoT systems without fully centralizing raw data. However, distributing data, computation, updates, and orchestration increases the attack surface and introduces coupled challenges across privacy, robustness, reliability, fairness, and governance.
By Long Dang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.