LSHFed: Robust and Communication-Efficient Federated Learning with Locally-Sensitive Hashing Gradient Mapping
Guanjie Cheng, Mengzhen Yang, Xinkui Zhao, Shuyi Yu, Tianyu Du, Yangyang Wu, Mengying Zhu, Shuiguang Deng · 2025
Read LSHFed: Robust and Communication-Efficient Federated Learning with Locally-Sensitive Hashing Gradient Mapping by Guanjie Cheng, Mengzhen Yang, Xinkui Zhao, Shuyi Yu, Tianyu Du, Yangyang Wu, Mengying Zhu, Shuiguang Deng (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01296.txt.
Description
Guanjie Cheng and colleagues harden federated learning, where many devices train one shared model without sending raw data to a center. The setup invites attack: gradients can leak private information, and malicious participants can poison the model. Existing defenses tend to be expensive or imprecise.
Their framework, LSHFed, checks gradients by hashing them. A locally sensitive hashing scheme with multiple hyperplanes compresses each high dimensional gradient into a compact binary code, irreversible by design. Malicious gradients can then be detected and filtered from their codes alone, cutting both privacy leakage and transmission cost. Experiments show the model stays accurate even when up to half the participants are colluding adversaries, while gradient verification communication drops by up to a thousand fold.
Catalog record
| CREATOR | Guanjie Cheng, Mengzhen Yang, Xinkui Zhao, Shuyi Yu, Tianyu Du, Yangyang Wu, Mengying Zhu, Shuiguang Deng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-01296