Abstract ：

In response to security and privacy in the integration of blockchain technology with federated learning (FL), comprehensive review and analysis of the relevant methods for empowering FL with blockchain are provided. First, FL and blockchain were elucidated separately, and on the basis of this, the state-of-the-art general architectures for blockchain-enabled FL were summarized. Second, the progress in security, privacy, incentives, and efficiency methods was investigated, and the advantages and disadvantages of each method were analyzed. Finally, the current issues in blockchain enabled FL were identified, and potential solutions were proposed, along with future prospects. © 2025 Beijing University of Technology. All rights reserved.

Keyword ：

efficiency data privacy incentive mechanism federated learning (FL) data security blockchain

Cite：

Copy from the list or Export to your reference management。

Abstract ：

针对区块链技术与联邦学习(federated learning,FL)结合后在安全、隐私等方面存在的问题,对区块链赋能FL中的相关方法进行综述与分析.首先,分别阐述了FL和区块链,并在此基础上总结了区块链赋能FL的前沿通用架构;其次,研究了目前安全、隐私、激励以及效率方法的进展,分析了各方法的优缺点;最后,指出了区块链赋能FL目前存在的问题,提出了解决方案,并进行了展望.

Keyword ：

FL) 效率 区块链 数据安全 激励机制 联邦学习(federated learning 数据隐私

Cite：

Copy from the list or Export to your reference management。

Abstract ：

高空平台(high altitude platform,HAP)技术与多接入边缘计算(multi-access edge computing,MEC)技术的结合将MEC服务器部署区域由地面扩展到空中,打破传统地面MEC网络的局限性,为用户提供无处不在的计算卸载服务。针对基于HAP的MEC计算卸载研究进行综述,首先,从HAP计算节点的优势、网络组成部分、网络结构、主要挑战及其应对技术4个方面分析基于HAP的MEC网络;其次,分别从图论、博弈论、机器学习、联邦学习等理论的角度对基于HAP的MEC计算卸载算法进行横向分析和纵向对比;最后,指出基于HAP的MEC计算卸载技术目前存在的问题,并对该技术的未来研究方向进行展望。

Keyword ：

机器学习 计算卸载 MEC) 多接入边缘计算(multi-access edge computing HAP) 高空平台(high altitude platform 博弈论 图论

Cite：

Copy from the list or Export to your reference management。

Abstract ：

目前，基于联邦深度强化学习的微网（MG）能量管理研究未考虑多类型能量转换与MG间电量交易的问题，同时，频繁交互模型参数导致通信时延较大。基于此，以一种包含风、光、电、气等多类型能源的MG为研究对象，构建了支持MG间电量交易和MG内能量转换的能量管理模型，提出基于正余弦算法的联邦竞争深度Q网络学习算法，并基于该算法设计了计及能量交易与转换的多MG能量管理与优化策略。仿真结果表明，所提能量管理策略在保护数据隐私的前提下，能够得到更高奖励且最大化MG经济收益，同时降低了通信时延。

Keyword ：

能量管理 微网(MG) 正余弦算法 联邦学习 竞争深度Q网络

Cite：

Copy from the list or Export to your reference management。

Abstract ：

针对模型异构和代理数据稀缺问题,提出模型异构的联邦学习入侵检测(model heterogeneous federated learning for intrusion detection,MHFL-ID)框架。首先,MHFL-ID根据模型异同对节点进行分组,将结构相同的模型分到同一组;其次,在组内采用以组长为中心的同构聚合方法,根据目标函数投影值选取组长,并引导组内节点的优化方向以提升全组模型能力;最后,在组间采用基于知识蒸馏的异构聚合方法,不需要代理数据就能用局部平均软标签和全局软标签传递异构模型中的知识。在NSL-KDD和UNSW-NB15这2个数据集上进行对比实验,与当前先进方法相比,MHFL-ID框架及所提方法能有效解决联邦学习中模型异构聚合的问题,在准确率方面也取得了较好结果。

Keyword ：

异构聚合 知识蒸馏 入侵检测 多目标 联邦学习 模型异构

Cite：

Copy from the list or Export to your reference management。

Abstract ：

为了解决物联网设备在资源受限和数据非独立同分布(Non-IID)时出现的训练效率低、模型性能差的问题,提出了一种个性化剪枝联邦学习框架用于物联网的入侵检测.首先,提出了一种基于通道重要性评分的结构化剪枝策略,该策略通过平衡模型的准确率与复杂度来生成子模型下发给资源受限客户端.其次,提出了一种异构模型聚合算法,对通道采用相似度加权系数进行加权平均,有效降低了Non-IID数据在模型聚合中的负面影响.最后,网络入侵数据集BoT-IoT上的实验结果表明,相较于现有方法,所提方法能显著降低资源受限客户端的时间开销,处理速度提升20.82%,并且在Non-IID场景下,入侵检测的准确率提高0.86%.

Keyword ：

联邦学习 非独立同分布 入侵检测 模型剪枝

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Aiming at the problem of model heterogeneity and agent data scarcity, a model heterogeneous federated learning for intrusion detection (MHFL-ID) framework was proposed. First, MHFL-ID groups nodes according to model similarities and differences, that is, models with the same structure were grouped into the same group. Second, the group leader centered isomorphic aggregation method was used to select the group leader according to the projection value of the objective function and guide the optimization direction of the nodes in the group to enhance the modeling capability of the whole group model. Finally, the heterogeneous aggregation method based on knowledge distillation was used between groups to transfer knowledge in heterogeneous models with local average soft label and global soft label without proxy data. Comparative experiments on two datasets, NSL-KDD and UNSW-NB15, show that MHFL-ID framework and the proposed method can effectively solve the problem of heterogeneous model aggregation in federated learning and achieve better results in terms of accuracy. © 2024 Beijing University of Technology. All rights reserved.

Keyword ：

intrusion detection multi-objective heterogeneous aggregation distillation of knowledge model heterogeneous federated learning

Cite：

Copy from the list or Export to your reference management。

Abstract ：

目前,随着人工智能研究的进步,人工智能被大规模采用,数据监管等领域的需求也促使人们对隐私保护的认识和关注越来越多,这促进了联邦学习(federated learning,FL)框架的流行.但现有的FL难以应对异构问题以及用户的个性化需求.为了应对上述问题,研究了个性化联邦学习(personalized federated learning,PFL)的相关方法并提出了展望.列举了FL的框架并指出了FL的不足,在FL场景的基础上,引出PFL的研究动机对PFL中的统计异构、模型异构、通信异构、设备异构进行分析并提出可行性方案;将PFL中的客户端选择、知识蒸馏等个性化算法分类并分析各自的创新与不足.最后,对PFL的未来研究方向进行了展望.

Keyword ：

数据监管 异构问题 隐私保护 个性化联邦学习(PFL)

Cite：

Copy from the list or Export to your reference management。

Abstract ：

To overcome the limitation of the Federated Learning (FL) when the data and model of each client are all heterogenous and improve the accuracy, a personalized Federated learning algorithm with Collation game and Knowledge distillation (pFedCK) is proposed. Firstly, each client uploads its soft-predict on public dataset and downloads the most correlative of the k soft-predict. Then, the Shapley Value (SV) from collation game is applied to measure the multi-wise influences among clients and their marginal contribution to others on personalized learning performance is quantified. Lastly, each client identify it’s optimal coalition and then the Knowledge Distillation (KD) is used to local model and local training is conduct on the privacy dataset. The results show that compared with the state-of-the-art algorithm, this approach can achieve superior personalized accuracy and can improve by about 10%. © 2023 Science Press. All rights reserved.

Keyword ：

Knowledge Distillation (KD) Personalized Federated Learning(PFL) Heterogeneity Collation game

Cite：

Copy from the list or Export to your reference management。

Abstract ：

In federated learning (FL), the distribution of data in clients is always heterogeneous, which makes the unified model trained in FL unable to meet the demand of each client. To combat this issue, a personalized federated learning algorithm with meta learning and knowledge distillation is proposed, in which the knowledge distillation and meta-learning with FL and incorporating the personalization are combined into the training of FL. In each global iteration, the global model (teacher model) update itself according to the feedback from the local model (student model) during the knowledge distillation. Therefore, each client can obtain a better personalized model. Simulation results show that compared with the existing personalized algorithms, the proposed algorithm can achieve a better compromise between global accuracy and personalization accuracy while improving the personalization accuracy. © 2023 Beijing University of Posts and Telecommunications. All rights reserved.

Keyword ：

meta-learning personalization federated learning knowledge distillation

Cite：

Copy from the list or Export to your reference management。