Federated Learning System Research Articles

Embedded Implementation and Evaluation of Deep Neural Network of Federated Learning

Compared with traditional distributed machine learning, federated learning (or joint learning) enables multiple computing nodes to cooperate and train a shared machine learning model without transmitting original data. At present, the research work of federated learning mainly focuses on the theoretical method, and the system implementation is less, and only for the text data or simple image such as medical institution information sharing, handwriting font recognition and other simple neural network applications. Aiming at more complex deep neural networks, this project implements a multi-node federated learning system on embedded device, and evaluates its key performance indicators such as training accuracy, delay and loss. The research method mainly uses embedded computer both as client and server, adjusts and groups the Visdrone datasets as training samples, and then trains the model on the client based on YOLOv4 algorithm, realizes the encrypted transmission of information through TCP protocol, and achieves the aggregation update of the model on the server with FedAvg algorithm.

Regulated Subspace Projection Based Local Model Update Compression for Communication-Efficient Federated Learning

Despite high utility in distributed networks, federated learning entails enormous communication overhead due to the requirement of trained model exchange at every global iteration. When the communication resources are limited, as in wireless environments, learning performance can be severely degraded by the communication overhead. On this account, communication efficiency is one of the primary concerns in federated learning. In this paper, we put forth a communication-efficient federated learning system based on the projection of local model updates. Leveraging the correlation of consecutive local model updates, we devise a novel local model update compression scheme based on the projection onto the selected subspace. Furthermore, to avoid error propagation over global iterations and thus improve learning performance, we also develop novel criteria for deciding whether to compress the local model updates or not. The convergence of the proposed algorithm is also mathematically proved by deriving an upper bound on the mean square error of the global parameter. The merits of the proposed algorithm over the state-of-the-art benchmark schemes are verified by various simulations.

Hierarchical Clustering-based Personalized Federated Learning for Robust and Fair Human Activity Recognition

Currently, federated learning (FL) can enable users to collaboratively train a global model while protecting the privacy of user data, which has been applied to human activity recognition (HAR) tasks. However, in real HAR scenarios, deploying an FL system needs to consider multiple aspects, including system accuracy, fairness, robustness, and scalability. Most existing FL frameworks aim to solve specific problems while ignoring other properties. In this paper, we propose FedCHAR, a personalized FL framework with a hierarchical clustering method for robust and fair HAR, which not only improves the accuracy and the fairness of model performance by exploiting the intrinsically similar relationship between users but also enhances the robustness of the system by identifying malicious nodes through clustering in attack scenarios. In addition, to enhance the scalability of FedCHAR, we also propose FedCHAR-DC, a scalable and adaptive FL framework which is featured by dynamic clustering and adapting to the addition of new users or the evolution of datasets for realistic FL-based HAR scenarios. We conduct extensive experiments to evaluate the performance of FedCHAR on seven datasets of different sizes. The results demonstrate that FedCHAR could obtain better performance on different datasets than the other five state-of-the-art methods in terms of accuracy, robustness, and fairness. We further validate that FedCHAR-DC exhibits satisfactory scalability on three large-scale datasets regardless of the number of participants.

LAFED: A lightweight authentication mechanism for blockchain-enabled federated learning system

Federated learning, as an emerging distributed machine learning technology, can use cross-device data to train a usable and secure shared model under the premise of protecting data privacy. However, the existing federated learning usually uploads the intermediate parameters to the central server to achieve model aggregation, which will cause significant privacy leakage. Recently, blockchain technology has become a research hotspot due to its advantages of decentralized and non-tampering features, providing new ideas for the realization of security certification for federated learning. However, blockchain-enabled federated learning also faces the following two challenges: (1) the identity authentication relies on the central server being fully trusted and the computation cost is heavy; (2) center-less authentication faces the challenges of efficiency and privacy leakage. To solve the above challenges, we propose a lightweight authentication mechanism for blockchain-enabled federated learning system, named LAFED. The innovations of LAFED are three-fold: (1) a lightweight authentication framework for blockchain-enabled federated learning; (2) a flexible consensus algorithm with zero-knowledge proof to verify the identity of each participant; (3) an adaptive model aggregation algorithm based on the model quality and node contribution to improve the performance. Extensive experimental results demonstrate that the proposed LAFED can achieve lightweight authentication while ensuring a high model accuracy.

Aggregation Service for Federated Learning: An Efficient, Secure, and More Resilient Realization

Federated learning has recently emerged as a paradigm promising the benefits of harnessing rich data from diverse sources to train high quality models, with the salient features that training datasets never leave local devices. Only model updates are locally computed and shared for aggregation to produce a global model. While federated learning greatly alleviates the privacy concerns as opposed to learning with centralized data, sharing model updates still poses privacy risks. In this paper, we present a system design which offers efficient protection of individual model updates throughout the learning procedure, allowing clients to only provide obscured model updates while a cloud server can still perform the aggregation. Our federated learning system first departs from prior works by supporting lightweight encryption and aggregation, and resilience against drop-out clients with no impact on their participation in future rounds. Meanwhile, prior work largely overlooks bandwidth efficiency optimization in the ciphertext domain and the support of security against an actively adversarial cloud server, which we also fully explore in this paper and provide effective and efficient mechanisms. Extensive experiments over several benchmark datasets (MNIST, CIFAR-10, and CelebA) show our system achieves accuracy comparable to the plaintext baseline, with practical performance.

ARMOR: Differential Model Distribution for Adversarially Robust Federated Learning

In this work, we formalize the concept of differential model robustness (DMR), a new property for ensuring model security in federated learning (FL) systems. For most conventional FL frameworks, all clients receive the same global model. If there exists a Byzantine client who maliciously generates adversarial samples against the global model, the attack will be immediately transferred to all other benign clients. To address the attack transferability concern and improve the DMR of FL systems, we propose the notion of differential model distribution (DMD) where the server distributes different models to different clients. As a concrete instantiation of DMD, we propose the ARMOR framework utilizing differential adversarial training to prevent a corrupted client from launching white-box adversarial attack against other clients, for the local model received by the corrupted client is different from that of benign clients. Through extensive experiments, we demonstrate that ARMOR can significantly reduce both the attack success rate (ASR) and average adversarial transfer rate (AATR) across different FL settings. For instance, for a 35-client FL system, the ASR and AATR can be reduced by as much as 85% and 80% over the MNIST dataset.

Uplink and Downlink Decoupled 5G/B5G Vehicular Networks: A Federated Learning Assisted Client Selection Method

Owning to the powerful support of 5G/B5G technologies, it is promising that, in a vehicular network, the ever-increasing demands for artificial intelligence applications along with privacy concerns can be fulfilled by introducing federated learning (FL) into vehicular networks. However, in such a 5G/B5G cellular-based vehicular network, distributed clients, i.e., vehicles, usually suffer from intermittent connectivity to the uplink base station (BS) due to the small cell networking and their mobility, and thereby the FL execution time and the learning performance cannot be guaranteed. In order to ensure a required execution time and the FL performance, we propose to investigate the client selection schemes in the uplink-downlink decoupled 5G/B5G networks. Due to the bursty feature of uplink transmissions, we first incorporate a flexible delay analytical tool, i.e., martingale theory, to characterize the access delay of each vehicle, and represent it in the form of delay violation probability. Then, given an energy budget, adjustable parameters are optimized to maximize the global learning performance. Supposing that all the clients participate in the FL and they can outperform the systematic FL performance requirement, the client selection methods under different criteria are further investigated. Especially, we propose a low-complexity algorithm to select clients in terms of enhancing energy efficiency. Numerical results are provided to demonstrate the effectiveness of the proposed client selection method in both guaranteeing the FL execution time and meeting the systematic learning performance. The energy efficiency of the vehicular FL system can also be enhanced by employing the proposed method.

A Scalable and Transferable Federated Learning System for Classifying Healthcare Sensor Data.

With the development of Internet of Medical Things, massive healthcare sensor data (HSD) are transmitted in the Internet, which faces various security problems. Healthcare data are sensitive and important for patients. Automatic classification of HSD has significant value for protecting the privacy of patients. Recently, the edge computing-based federated learning has brought new opportunities and challenges. It is difficult to develop a lightweight HSD classification system for edge computing. In particular, the classification system should consider the dynamic characteristics of HSD, e.g., the change of data distributions and the appearance of initially unknown classes. To solve these problems, the paper proposes a scalable and transferable classification system, called SCALT. It is a one-classifier-per-class system based on federated learning. It comprises a one-dimensional convolution-based network for feature extraction, and an individual mini-classifier for each class. It is easy to be scaled when new class appears since only a mini-classifier will be trained. The feature extractor is updated only when it is transferred to a new task. SCALT has a parameter protection mechanism, which can avoid catastrophic forgetting in sequential HSD classification tasks. We conduct comprehensive experiments to evaluate SCALT on three different physiological signal datasets: Electrocardiogram, Electroencephalogram and Photoplethysmograph. The accuracies on the three datasets are 98.65%, 91.10% and 89.93% respectively, which are higher than the compared state-of-the-art works. At last, an application of applying SCALT to protect the privacy of patients is presented.

Blockchain-Based Decentralized Federated Learning Method in Edge Computing Environment

In recent years, federated learning has been able to provide an effective solution for data privacy protection, so it has been widely used in financial, medical, and other fields. However, traditional federated learning still suffers from single-point server failure, which is a frequent issue from the centralized server for global model aggregation. Additionally, it also lacks an incentive mechanism, which leads to the insufficient contribution of local devices to global model training. In this paper, we propose a blockchain-based decentralized federated learning method, named BD-FL, to solve these problems. BD-FL combines blockchain and edge computing techniques to build a decentralized federated learning system. An incentive mechanism is introduced to motivate local devices to actively participate in federated learning model training. In order to minimize the cost of model training, BD-FL designs a preference-based stable matching algorithm to bind local devices with appropriate edge servers, which can reduce communication overhead. In addition, we propose a reputation-based practical Byzantine fault tolerance (R-PBFT) algorithm to optimize the consensus process of global model training in the blockchain. Experiment results show that BD-FL effectively reduces the model training time by up to 34.9% compared with several baseline federated learning methods. The R-PBFT algorithm can improve the training efficiency of BD-FL by 12.2%.

Personalized federated learning with model interpolation among client clusters and its application in smart home

The proliferation of high-performance personal devices and the widespread deployment of machine learning (ML) applications have led to two consequences: the volume of private data from individuals or groups has exploded over the past few years; and the traditional central servers for training ML models have experienced communication and performance bottlenecks in the face of massive amounts of data. However, this reality also provides the possibility of keeping data local for ML training and fusing models on a broader scale. As a new branch of ML application, Federated Learning (FL) aims to solve the problem of multi-party joint learning on the premise of protecting personal data privacy. However, due to the heterogeneity of devices, including network connection, network bandwidth, computing resources, etc., it is unrealistic to train, update and aggregate models in all devices in parallel, while personal data is often not independent and identically distributed (Non-IID) due to multiple reasons. This reality poses a challenge to the speed and convergence of FL. In this paper, we propose the pFedCAM algorithm, which aims to improve the robustness of the FL system to device heterogeneity and Non-IID data, while achieving some degree of federation model personalization. pFedCAM is based on the idea of clustering and model interpolation by classifying heterogeneous clients and performing FedAvg algorithm in parallel, and then combining them into personalized federated global models by inter-cluster model interpolation. Experiments show that the accuracy of pFedCAM improves 10.3% on Fashion-MNIST and 11.3% on CIFAR-10 compared to the benchmark in the case of Non-IID data. In the end, we applied pFedCAM in HomeProtect, a smart home privacy protection framework we designed, and achieved good practical results in the case of flame recognition.

Deep reinforcement learning‐based full‐duplex link scheduling in federated learning‐based computing for IoMT

SummaryThe rapid developments in mini‐hardware manufacturing and wireless network communications have enabled the Internet of Medical Things (IoMT) to provide continuous healthcare services over the Internet. Federated learning (FL) combined with blockchain technology has been a popular way to resolve privacy‐preserving data sharing in IoMT‐based wireless body area networks (WBANs), on the other side, communication payloads become much heavier than traditional healthcare sensor network, because central server should aggregate the model updates and orchestrate the training tasks. The high latency will lead to FL's low system efficiency. However, the existing studies on FL mainly focus on the system design and algorithm optimization, which ignore a critical problem of data transmission in the FL system. To improve the communication performance, we proposed a two‐tier scheduling algorithm in which a full‐duplex (FD) multiple access based scheduling algorithm is employed to improve channel utility and network throughput, and decrease the delay in tier II. A deep reinforcement learning (DRL) framework is used to generate the FD links between hubs and access points (APs) which jointly considers the channel state, fairness, and delay. Therefore, the DRL‐based FD Link Scheduling (R‐FDLS) algorithm is proposed. When the traffic volume is different or in various distribution scenarios, the evaluation results demonstrate that the proposed algorithm significantly improves the network communication quality, as well as has obvious advantages compared to several baselines.

Federating Medical Deep Learning Models from Private Jupyter Notebooks to Distributed Institutions

Deep learning-based algorithms have led to tremendous progress over the last years, but they face a bottleneck as their optimal development highly relies on access to large datasets. To mitigate this limitation, cross-silo federated learning has emerged as a way to train collaborative models among multiple institutions without having to share the raw data used for model training. However, although artificial intelligence experts have the expertise to develop state-of-the-art models and actively share their code through notebook environments, implementing a federated learning system in real-world applications entails significant engineering and deployment efforts. To reduce the complexity of federation setups and bridge the gap between federated learning and notebook users, this paper introduces a solution that leverages the Jupyter environment as part of the federated learning pipeline and simplifies its automation, the Notebook Federator. The feasibility of this approach is then demonstrated with a collaborative model solving a digital pathology image analysis task in which the federated model reaches an accuracy of 0.8633 on the test set, as compared to the centralized configurations for each institution obtaining 0.7881, 0.6514, and 0.8096, respectively. As a fast and reproducible tool, the proposed solution enables the deployment of a cross-country federated environment in only a few minutes.

Multi-Objective Optimization for Bandwidth-Limited Federated Learning in Wireless Edge Systems

This paper studies a bandwidth-limited federated learning (FL) system where the access point is a central server for aggregation and the energy-constrained user equipemnts (UEs) with limited computation capabilities (e.g., Internet of Things devices) perform local training. Limited by the bandwidth in wireless edge systems, only a part of UEs can participate in each FL training round. Selecting different UEs could affect the FL performance, and selected UEs need to allocate their computing resource effectively. In wireless edge FL systems, simultaneously accelerating FL training and reducing computing-communication energy consumption are of importance. To this end, we formulate a multi-objective optimization problem (MOP). In MOP, the model training convergence is difficult to calculate accurately. Meanwhile, MOP is a combinatorial optimization problem, with the high-dimension mix-integer variables, which is proved to be NP-hard. To address these challenges, a multi-objective evolutionary algorithm for the bandwidth-limited FL system (MOEA-FL) is proposed to obtain a Pareto optimal solution set. In MOEA-FL, an age-of-update-loss method is first proposed to transform the original global loss function into a convergence reference function. Then, MOEA-FL divides MOP into N single objective subproblems by the Tchebycheff approach and optimizes the subproblems simultaneously by evolving a population. Extensive experiments have been carried out on MNIST dataset and a medical case called TissueMNIST dataset for both the i.i.d and non-i.i.d data setting. Experimental results demonstrate that MOEA-FL performs better than other algorithms and verify the robustness and scalability of MOEA-FL.

Long-Term Privacy-Preserving Aggregation With User-Dynamics for Federated Learning

Privacy-preserving aggregation protocol is an essential building block in privacy-enhanced federated learning (FL), which enables the server to obtain the sum of users’ locally trained models while keeping local training data private. However, most of the work on privacy-preserving aggregation provides privacy guarantees for only one communication round in FL. In fact, as FL usually involves long-term training, i.e., multiple rounds, it may lead to more information leakages due to the dynamic user participation over rounds. In this connection, we propose a long-term privacy-preserving aggregation (LTPA) protocol providing both single-round and multi-round privacy guarantees. Specifically, we first introduce our batch-partitioning-dropping-updating (BPDU) strategy that enables any user-dynamic FL system to provide multi-round privacy guarantees. Then we present our LTPA construction which integrates our proposed BPDU strategy with the state-of-the-art privacy-preserving aggregation protocol. Furthermore, we investigate the impact of LTPA parameter settings on the trade-off between privacy guarantee, protocol efficiency, and FL convergence performance from both theoretical and experimental perspectives. Experimental results show that LTPA provides similar complexity to that of the state-of-the-art, i.e., an additional cost of around only 1.04X for a 100,000-user FL system, with an additional long-term privacy guarantee.

LoMar: A Local Defense Against Poisoning Attack on Federated Learning

Federated learning (FL) provides a high efficient decentralized machine learning framework, where the training data remains distributed at remote clients in a network. Though FL enables a privacy-preserving mobile edge computing framework using IoT devices, recent studies have shown that this approach is susceptible to poisoning attacks from the side of remote clients. To address the poisoning attacks on FL, we provide a <i>two-phase</i> defense algorithm called <inline-formula><tex-math notation="LaTeX">${\underline{Lo}cal\ \underline{Ma}licious\ Facto\underline{r}}$</tex-math></inline-formula> (LoMar). In phase I, LoMar scores model updates from each remote client by measuring the relative distribution over their neighbors using a kernel density estimation method. In phase II, an optimal threshold is approximated to distinguish malicious and clean updates from a statistical perspective. Comprehensive experiments on four real-world datasets have been conducted, and the experimental results show that our defense strategy can effectively protect the FL system. Specifically, the defense performance on Amazon dataset under a label-flipping attack indicates that, compared with FG+Krum, LoMar increases the target label testing accuracy from <inline-formula><tex-math notation="LaTeX">$96.0\%$</tex-math></inline-formula> to <inline-formula><tex-math notation="LaTeX">$98.8\%$</tex-math></inline-formula> , and the overall averaged testing accuracy from <inline-formula><tex-math notation="LaTeX">$90.1\%$</tex-math></inline-formula> to <inline-formula><tex-math notation="LaTeX">$97.0\%$</tex-math></inline-formula> .

Efficient and Secure Federated Learning With Verifiable Weighted Average Aggregation

Federated learning allows a large number of participants to collaboratively train a global model without sharing participant's local data. Participants train local models with their local data and send gradients to the cloud server for aggregation. Unfortunately, as a third party, the cloud server cannot be fully trusted. Existing research has shown that a compromised cloud server can extract sensitive information of participant's local data from gradients. In addition, it can even forge the aggregation result to corrupt the global model without being detected. Therefore, in a secure federated learning system, both the privacy and aggregation correctness of the uploaded gradients should be guaranteed. In this article, we propose a secure and efficient federated learning scheme with verifiable weighted average aggregation. By adopting the masking technique to encrypt both weighted gradients and data size, our scheme can support the privacy-preserving weighted average aggregation of gradients. Moreover, we design the verifiable aggregation tag and propose an efficient verification method to validate the weighted average aggregation result, which greatly improves the performance of the aggregation verification. Security analysis shows that our scheme is provably secure. Extensive experiments demonstrate the efficiency of our scheme compared with the state-of-the-art approaches.

Visual Object Detection for Privacy-Preserving Federated Learning

Visual object detection is a computer vision technique based on deep learning. However, due to privacy issues, it is highly challenging to build an object detection model according to the current methods on the centrally stored training dataset. Federated learning is an effective approach to solving the challenge of training data collection by using distributed training. We propose FedVisionBC, a blockchain-based federated learning system for visual object detection that addresses the challenges of single point of failure, model poisoning attacks, and membership inference attacks in traditional federated learning. In the FedVisionBC system, we set up an aggregation node and a verification node instead of a central server to solve the single point of failure problem. We establish a security mechanism that uses encryption techniques, verification nodes, and smart contracts to resist model poisoning attacks. Experimental results show that FedVisionBC can accomplish the object detection task when the percentage of malicious clients is less than 60%. We also propose a new algorithm, ADPFedAvg, to prevent membership inference attacks, which relies on user-level differential privacy technology and the federal average algorithm. Experimental results indicate that ADPFedAvg can achieve a large-scale visual object detection model with differential privacy protection, while only a negligible cost in predictive accuracy.

IoT: A Decentralized Trust Management System Using Blockchain-Empowered Federated Learning

We propose a decentralized medical trust management system using blockchain-based federated learning for large-scale Internet of Things (IoT) systems. The proposed system enables health institutions to share data without revealing the privacy of data owners. Health institutions form coalitions and the leader of each coalition is elected based on the proposed proof-of-trust collaboration (PoTC) consensus protocol. The PoTC consensus protocol is based on a weight difference game where trust scores, trust consistency value, and trust deviation are factors used for evaluating nodes in the blockchain. The trust of a node is obtained either through direct trust or recommended trust evaluations. Each leader elects an aggregator who has the most credibility to manage the proposed federated learning system. The leaders become the federated clients as well as validators while the aggregator is the federated server. To ensure the decentralization of nodes, a consortium blockchain is employed. Extensive simulations are performed, which show that the proposed system not only demonstrates scalability and credibility without compromising the accuracy, convergence, and resilience properties against malicious attackers but also outperforms existing trust management systems. A security analysis is also conducted, which shows that the proposed system is robust against trust-related attacks.

ClusterFL: A Clustering-based Federated Learning System for Human Activity Recognition

Federated Learning (FL) has recently received significant interest, thanks to its capability of protecting data privacy. However, existing FL paradigms yield unsatisfactory performance for a wide class of human activity recognition (HAR) applications, since they are oblivious to the intrinsic relationship between data of different users. We propose ClusterFL, a clustering-based federated learning system that can provide high model accuracy and low communication overhead for HAR applications. ClusterFL features a novel clustered multi-task federated learning framework that minimizes the empirical training loss of multiple learned models while automatically capturing the intrinsic clustering relationship among the nodes. We theoretically prove the convergence of proposed FL framework for non-convex and strongly convex models and provide the guidance on selection of hyper-parameters for achieving such convergence. Based on the learned cluster relationship, ClusterFL can efficiently drop the nodes that converge slower or have little correlations with others in each cluster, significantly speeding up the convergence while maintaining the accuracy performance. We evaluate the performance of ClusterFL on an NVIDIA edge testbed using four new HAR datasets collected from 145 users. The results show that ClusterFL outperforms several state-of-the-art FL paradigms in terms of overall accuracy and can save more than 50% communication overhead.

NttpFL: Privacy-Preserving Oriented No Trusted Third Party Federated Learning System Based on Blockchain

In federated learning, multiple parties may use their data to cooperatively train a model without exchanging raw data. Federated learning protects the privacy of users to a certain extent. However, model parameters may still expose private information. Moreover, existing encrypted federated learning systems need a trusted third party to generate and distribute key pairs to connected participants, making them unsuitable for federated learning and vulnerable to security risks. To mitigate these issues, we propose a privacy-preserving oriented no trusted third party federated learning system based on blockchain (NttpFL). The initiator of the federated learning task and the partners negotiate keys through the conference key agreement and do not need to distribute keys through a trusted third party. We design a double-layer encryption mechanism to ensure privacy. Partners cannot obtain any private information other than their information. The decentralized nature of blockchain suits our system. In addition, blockchain makes the entire process transparent and traceable and avoids the single node failure problem. Experimental results confirm that the proposed method significantly reduces the communication costs and computational complexity compared to existing encrypted federated learning without compromising the performance and security.