MISA: Unveiling the Vulnerabilities in Split Federated Learning
Not on the map. Why?
Abstract (arXiv)
textitFederated learning (FL) and textitsplit learning (SL) are prevailing distributed paradigms in recent years. They both enable shared global model training while keeping data localized on users' devices. The former excels in parallel execution capabilities, while the latter enjoys low dependence on edge computing resources and strong privacy protection. textitSplit federated learning (SFL) combines the strengths of both FL and SL, making it one of the most popular distributed architectures. Furthermore, a recent study has claimed that SFL exhibits robustness against poisoning attacks, with a fivefold improvement compared to FL in terms of robustness. In this paper, we present a novel poisoning attack known as MISA. It poisons both the top and bottom models, causing a textbfunderlinemisalignment in the global model, ultimately leading to a drastic accuracy collapse. This attack unveils the vulnerabilities in SFL, challenging the conventional belief that SFL is robust against poisoning attacks. Extensive experiments demonstrate that our proposed MISA poses a significant threat to the availability of SFL, underscoring the imperative for academia and industry to accord this matter due attention.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 12 of its 19 references that are in the map. First posted on arXiv 18 Dec 2023.
Builds on
12 of 19 references.
References · 12
- 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
- 2018Byzantine-Robust Distributed Learning: Towards Optimal Statistical Ratescited by 711
- 2019Local Model Poisoning Attacks to Byzantine-Robust Federated Learningcited by 493
- 2018Split learning for health: Distributed deep learning without sharing raw patient datacited by 395
- 2020SplitFed: When Federated Learning Meets Split Learningcited by 310
- 2019A Little Is Enough: Circumventing Defenses For Distributed Learningcited by 288
- 2019Fall of Empires: Breaking Byzantine-tolerant SGD by Inner Product Manipulationcited by 141
- 2021Challenges and Approaches for Mitigating Byzantine Attacks in Federated Learningcited by 47
- 2022Shielding Federated Learning: Robust Aggregation with Adaptive Client Selectioncited by 18
- 2023A Four-Pronged Defense Against Byzantine Attacks in Federated Learningcited by 9
- 2022Security Analysis of SplitFed Learningcited by 5
- 2023Analyzing the vulnerabilities in SplitFed Learning: Assessing the robustness against Data Poisoning Attackscited by 4
Built on by
Citing papers per year: 2024: 1; 2025: 2.
3 papers.