Earticle

최근 인터넷이 발달함에 따라 공기 청정기, 가정용 CCTV, 냉장고 등 가정 내 IoT(Internet Of Things)기기들이 많이 개발되고, 사용되고 있다. 현대의 빠르게 확장되는 IoT 생태계에서 데이터의 분산은 개인 장치 및 센서로부터 비롯되어 중앙 집중식 학습 모델의 적용을 어렵게 만들고 있다. 이에 대한 대안으로써 FL(Federated Learning,연합 학습)은 모델 학습을 로컬 디바이스로 분산 시켜 개인 데이터를 중앙 서버로 전송 하지 않고도 학습이 가능하게 한다. 본 논문에서는 FL의 핵심 원리를 바탕으로 IoT 환경에서 데이터 보안 및 프라이버시 문제를 확인한다, 또한 실제 응용 분야에서의 적용 사례 및 성공 사례를 살펴보며, FL이 실제 IoT환경에서 어떻게 적용 되고 있는지를 분석한다.마지막으로 효과적인 FL구현을 위한 향후 연구 방향을 소개한다.

With the recent development of the Internet, various household IoT (Internet of Things) devices such as air purifiers, home CCTV, and refrigerators have been extensively developed and adopted. In the rapidly expanding modern IoT ecosystem, the decentralization of data originating from personal devices and sensors poses challenges for the application of centralized learning models. As an alternative, Federated Learning (FL) enables model training to be distributed to local devices, allowing learning without transmitting personal data to a central server. This paper explores the core principles of FL in the context of IoT, examining issues related to data security and privacy. Furthermore, it delves into real-world applications and success stories, analyzing how FL is currently being applied in practical IoT environments. Lastly, the paper introduces future research directions to facilitate effective implementation of FL.

Hospitals have accumulated large amounts of patient data, with each hospital’s data having unique characteristics and distributions. By leveraging this vast amount of data for machine learning, we can develop predictive models, such as those for predicting long-term outcomes in ischemic stroke patients and provide valuable information for treatment decisions. However, data privacy concerns prevent hospital data from being put together on a centralized server. This study investigates the applicability of federated learning for predicting long-term outcomes in ischemic stroke patients using data from Hallym University Sacred Heart Hospital in Pyeongchon and Hallym University Sacred Heart Hospital in Chuncheon. Patient outcomes are defined as favorable if the modified Rankin Scale (mRS) score is 0-2 and poor if the mRS score is 3-6. There are two tasks: one predicting patient outcomes at 3 months after stroke and the other predicting patient outcomes at 1 year after stroke. A simple deep neural networks model is used for implementation of the prediction model and the federated learning environment. In conclusion, the federated learning models using basic FedAVG and weighted averaging FedAVG achieved 99.4%-99.9% performance of traditional centralized learning models.

Phishing attacks increasingly exploit mobile platforms and target users communicating in lowresource or code-mixed languages, posing challenges to traditional centralized detection systems. This study proposes hybrid knowledge distillation and federated learning framework for real-time, on-device phishing detection. The approach integrates a fine-tuned XLM-RoBERTa "teacher" model with a compact MobileBERT "student" model, distilled to achieve near teacher-level performance while enabling efficient offline inference. The distilled MobileBERT model, converted to ONNX for platform portability, achieved a fivefold reduction in size while preserving 98% of the original model’s accuracy. We conducted zero-shot evaluations on Korean, Spanish, and Turkish datasets to guarantee crosslinguistic robustness, and consistently obtained good accuracy and recall rates. Moreover, privacy-preserving updates were made possible using a federated learning simulation, which permits decentralized model enhancements without data exchange. The suggested architecture offers a cost-effective, expandable, and privacy-conscious approach to phishing detection in scenarios with limited connection and linguistic diversity.

분산형 농업 IoT 애플리케이션에서 데이터 프라이버시 문제는 스마트 온실 시스템이 생성하는 방대한 양의 민감한 환경 및 작물 관련 데이터로 인해 제기된다. 본 연구는 온실 환경에서의 데이터 보호를 개선하기 위해 차등 프라이버시를 통합한 프라이버시 강화 FL 아키텍처를 제안한다. 이 접근법은 특히 라즈베리 파이와 같은 제한된 자원을 가진 에지 디바이스를 위해 설계했다. 실제 온실 환경 데이터셋을 대상으로 한 실험에 따르면, 중앙 집중식 훈련은 최대 정확도(92.4%)를 달성하지만 프라이버시 보호 기능이 전혀 제공되지 않는다. 정확도, 프라이 버시 예산, 시스템 효율성 간의 균형 잡힌 절충점을 달성하기 위해 제안된 계층적 FL+DP 전략은 모델 업데이트를 크게 안정화하고 통신 오버헤드를 낮춘다. 이러한 결과는 높은 데이터 기밀성과 정확한 예측이 필요한 프라이버시 민감형 스마트 온실 애플리케이션에 FLDP 프레임워크가 적합함을 보여주었다.

Concerns with data privacy in dispersed agricultural IoT applications are raised by the massive amounts of sensitive environmental and crop-related data generated by smart greenhouse systems. In order to improve data protection in greenhouse settings, this study suggests a privacy -enhanced federated learning architecture that incorporates differential privacy. In order to achieve a balanced trade-off between accuracy, privacy budget, and system efficiency, the suggested hierarchical FL+DP strategy significantly stabilizes model updates and lowers communication overhead. The suggested model shows a favorable privacy-accuracy trade-off when compared to ordinary federated learning, achieving up to 88.7% prediction accuracy under moderate noise circumstances (σ = 0.3–0.5) while offering explicit differential privacy guarantees. These findings show that the FLDP framework is a good fit for privacy-sensitive smart greenhouse applications that need high data secrecy and accurate prediction.

연방학습은 환자의 원시 데이터를 교환하지 않고 여러 의료기관간 협력적 모델훈련을 가능하게 함으로써 의료 AI의 개인정보 보호 문제를 해결한다. 그러나 기존 연방학습은 여전히 기울기 노출로 인한 개인정보 유출 위 험이 존재한다. 본 연구는 데이터와 모델 매개변수를 보호하기 위해 차등 프라이버시(DP)와 동형 암호화(HE)를 결합한 프라이버시 강화연방학습 프레임워크를 제안한다. 동형암호는 암호화된 연산을 통해 통신을 암호화하는 반 면, 차등 프라이버시는 기울기에 제어된 노이즈를 추가한다. 실제 의료 데이터셋을 활용한 질병분류 실험결과, 제안 한 모델은 우수한 프라이버시(ε = 1.2)와 적당한 통신비용으로 90.7%의 정확도를 보였다. 안전한 의료 애플리케이 션을 위해 본 모델은 효율성, 프라이버시, 성능간의 적절한 균형을 제공한다.

Federated Learning (FL) addresses privacy concerns in medical AI by facilitating cooperative model training across several healthcare organizations without exchanging patients' raw data. However, traditional FL still runs the danger of privacy leaking due to gradient exposure. In order to protect data and model parameters, this research aims to provide a privacy-enhanced FL framework that combines Differential Privacy (DP) with Homomorphic Encryption (HE). While HE encrypts communication with encrypted computation, DP adds controlled noise to gradients. Experiments on a real medical dataset for disease classification demonstrate that the suggested FL + DP + HE strategy achieves 90.7% accuracy with good privacy (ε = 1.2) and modest communication cost. For safe healthcare applications, the model provides a good balance between efficiency, privacy, and performance.

연합학습(Federated Learning, FL)은 여러 참여자가 원본 데이터를 직접 교환하지 않고 분산된 환경에서 공동으로 모델을 학습함으로써 개인정보를 보호하고 효율적 협업을 달성할 수 있는 기술로 주목받고 있다. 그러나 참여자의 자발적 협력에 의존하는 특성상, 참여 유도와 기여 품질 확보를 위한 효과적인 인센티브 설계가 중요하 다. 이에 본 연구는 연합학습의 인센티브 메커니즘에 관한 최신 연구를 종합적으로 검토하고, Transformer 기반 토픽 모델링 기법인 BERTopic을 활용하여 연구 동향을 분석하였다. 분석 결과, 인센티브 유형, 행동적 가정, 평 가 전략, 블록체인 통합의 네 가지 차원으로 구성된 설계 프레임워크를 제시하였으며, 토큰 기반·평판 기반·하이브 리드형 인센티브 모델의 특성과 한계를 비교하였다. 마지막으로, 엣지 컴퓨팅, 헬스케어, 기관 간 협업 등 다양한 응용 분야에서의 시사점을 논의하며, 인센티브 메커니즘 연구가 기술·경제·행동과학이 융합된 다학제적 연구 영역 임을 확인하였다.

Federated Learning (FL) allows multiple participants to train shared models without exchanging raw data, protecting privacy but relying on voluntary participation. Effective incentives are therefore critical to ensure engagement and quality contributions. This review synthesizes recent developments, applying transformer-based topic modeling (BERTopic) to classify and compare incentive mechanisms. We propose a four-dimensional framework—covering incentive type, behavioral assumptions, evaluation strategy, and blockchain integration—and examine three primary models: token-based, reputation-based, and hybrid. Token incentives provide clear financial motivation but risk manipulation; reputation systems build trust but face cold-start and scalability issues; hybrids combine strengths but add complexity. We outline domain-specific implications for settings like edge computing, healthcare, and cross-organization collaboration, highlighting the interdisciplinary nature of incentive design.

실제 환경에서 연합학습을 수행하기 위해 많은 기기가 준비되어야 하고 다양한 상황들에 대한 고려가 필 요하다. 이는 높은 비용 소모를 야기하므로 실제 환경에서 연합학습을 구축하기 전에 시뮬레이션을 통해 성능 측 정을 해보는 과정이 필요하다. 또한 기존의 연합학습 방법은 개인이 소유하고 있는 다양한 기기들을 고려하지 않 는다는 한계를 가지고 있다. 이와 같은 한계를 극복하기 위해 본 논문에서는 연합학습에 사용되는 다양한 인자들을 파라미터화한 시뮬레이터를 구현하였고 개인이 소유한 다양한 기기가 활용된 학습이 고려될 수 있도록 Learning Agent 개념을 추가하였다. 제안한 시뮬레이터는 Windows 11, Python3.9의 환경에서 구현되었으며, 본 논문에서 가정한 상황에서 시뮬레이션 한 결과 Learning Agent를 사용한 경우, 평균적으로 전체 학습 시간은 1/2 가량 줄 어들었으며 상한 정확도에 수렴하기까지의 소요 시간은 1/3 가량 줄어듦을 확인하였다. 따라서 제안된 시뮬레이터 는 사용자의 다양한 기기를 활용한 연구에 활용될 수 있을 것이며 실제 환경에 연합학습을 구축하려는 상황에서 성능 측정을 위해 활용할 수 있는 도구가 될 것으로 기대한다.

To perform federated learning (FL) in real environments, many devices are required and various situations should be considered. This leads to high-cost consumption, so we need a process to measure performance through simulation before conducting FL in a real environment. In addition, the existing FL has a limitation in that it does not consider users’ devices which can be utilized for training. To overcome such limitation, in this paper, we implemented the FL simulator that parameterizes various factors in FL to consider diverse devices and situations. Also, the concept of Learning Agent was considered so that the simulator includes training by using users’ idle devices. The simulator was implemented on Windows 11 using Python 3.9. Using the simulator, we performed the various simulations in the assumed situations. As a result, on average, the total training time was halved and the time required to reach the saturation point of accuracy was reduced by one-third. These results showed that the proposed simulator can be used for researches using various users’ idle devices and measure performance before constructing FL systems in real environments.

연합학습은 데이터를 중앙 서버에 모아서 학습하는 중앙집중식 학습과 달리 개별 기기에서 모델 학습을 진행한 후 중앙 서버로 가중치들을 전송하는 분산식 학습 방식을 가진다. 따라서 개별 기기의 연산 성능과 통신 성 능 및 상황이 모델 학습에 큰 영향을 끼칠 수 있으므로 적절한 기기와 통신 방식을 선택해야 한다. 하지만 기존 연 합학습은 클라이언트 주위의 신뢰할 수 있는 가용 유휴 장치와 다양한 통신 및 다중 경로를 사용하지 않는 한계가 존재한다. 이를 극복하기 위해, 본 논문에서는 Learning agent 활용 및 다중 경로 기반 모델 분할 전송을 통한 연 합학습 개선 연구를 수행하였다. 시뮬레이션과 실제 장치를 활용한 실험을 통해 제안한 기법을 평가하였으며, 기존 연합학습과 비교하여 제안한 기법을 활용한 경우 학습 시간의 감소 및 학습 성능의 향상을 확인할 수 있었다.

Unlike centralized learning, federated learning has a distributed learning method that individual devices perform model training locally and then transmit weights to the central server. Therefore, the appropriate devices and communication methods should be selected as the computational and communication performance, and situation of individual devices can have a great impact on learning. However, the existing federated learning has limitations in not using various communications through multiple paths simultaneously and reliable and available idle devices around clients. To overcome this, in this paper, we conduct a study on improvement of federated learning using learning agent and multipath-based model split transmission. We evaluated the proposed technique through simulations and experiments using the real devices. The evaluation results show that the training time was reduced and the learning performance improved when the proposed technique was used compared to the existing method.

본 연구에서는 분산 컴퓨팅 및 개별 디바이스 활용을 통해 개인 정보 보호에 특화된 학습 방법인 연합학습 방법론을 기반으로, 모바일 내비게이션 애플리케이션에서 수집된 대규모의 운전자 데이터를 이용하여 경로 선택 예측 모델을 수립하는 방법에 대해 고찰한다. 경로 선택 모델링에서 활용될 수 있는 운전자 데이터의 전처리 및 분석 방법을 수립하고, 서포트벡터머 신(SVM) 및 다층 퍼셉트론(MLP)과 같이 기존에 널리 활용되는 학습 방법과 연합학습 방법의 성능과 특성을 비교한다. 분석 결과 연합학습을 통한 모델 성능은 중앙 서버 기반의 모델과의 비교에서 예측 정확도 측면의 차이가 거의 없는 것으로 나타났으나, 개별 데이터가 충분히 확 보되는 경우 연합학습 모델과 같은 개인화 모델의 성능이 개선될 수 있다는 점을 확인하였다. 연합학습 모델은 본 연구의 경로 선택 모델링 사례와 같이 모빌리티 부문의 데이터 프라이버 시 문제가 중요한 분야에서 대규모 데이터 처리를 필요로 하는 경우에 그 활용 가치가 매우 높을 것으로 기대된다.

The use of big data for transportation often involves using data that includes personal information, such as the driver's driving routes and coordinates. This study explores the creation of a route choice prediction model using a large dataset from mobile navigation apps using federated learning. This privacy-focused method used distributed computing and individual device usage. This study established preprocessing and analysis methods for driver data that can be used in route choice modeling and compared the performance and characteristics of widely used learning methods with federated learning methods. The performance of the model through federated learning did not show significantly superior results compared to previous models, but there was no substantial difference in the prediction accuracy. In conclusion, federated learning-based prediction models can be utilized appropriately in areas sensitive to privacy without requiring relatively high predictive accuracy, such as a driver's preferred route choice.

연합학습(Federated Learning)은 로컬데이터 기반의 학습 파라미터를 서버로 보내어 파라미터를 업데이트하는 방식의 분산 학습 방식으로써 로컬 데이터 유출을 방지하면서 모델을 공동 학습할 수 있는 구조를 제공한다. 연합학 습은 로컬 데이터의 유출은 방지하지만 파라미터 전송으로 인한 네트워크 부하와 파라미터에 대한 역복원 공격 (Inversion Attack)에 취약하다는 한계가 있다. 본 연구는 연합학습에서 파라미터 전송 효율성과 보안성을 향상시 키기 위해 확장 양자화 방식을 적용한 성능 분석을 다룬다. 기존의 gradient 클리핑, 차등 프라이버시, 양자화 기법 은 역복원 공격에 대한 보안성을 높이지만 전송량 절감 효과에는 제한적이다. 본 연구는 양자화 시 값의 범위를 확 장하여 양자화를 수행함으로써 유효 파라미터값을 줄이는 확장 양자화 기법을 적용하여 보았고, 다양한 확장 비율과 양자화 레벨에 따른 전송량 절감 및 역복원 공격 방어 성능을 실험적으로 분석했으며, 실험 결과, 확장 양자화는 약 간의 정확도 손실을 감수하는 대신 전송량을 효과적으로 줄이고 보안성을 향상시킬 수 있음을 확인하였다.

Federated Learning is distributed learning method in which locally trained parameters are sent to a server for parameter updates, enabling collaborative model training while preventing the leakage of local data. Although it protects local data from exposure, Federated Learning still suffers from limitations such as network overhead caused by parameter transmission and vulnerability to inversion attacks on the transmitted parameters. This study analyzes the performance of applying an extended quantization method to improve both parameter transmission efficiency and security in Federated Learning. While existing techniques such as gradient clipping, differential privacy, and standard quantization can enhance security against inversion attacks, their effectiveness in reducing transmission volume is limited. In this work, we applied an extended quantization technique that reduces the number of effective parameter values by expanding the quantization range, and we experimentally analyzed its impact on transmission reduction and defense against inversion attacks across various expansion ratios and quantization levels. Experiments show that extended quantization can effectively reduce transmission volume and improve security, with the cost of small loss in model accuracy.

병원 간 전자의무기록 (EHR) 시스템은 중앙 집중식 학습 최적화를 제한하는 개인정보 보호 및 업무량 다양성 문제에 직면해 있다. 본 논문에서는 기관의 자율성을 유지하면서 비용 예측 및 쿼리 실행 효율성을 향상시키는 연합 쿼리 최적화 프레임워크를 구축하고자 한다. 계획 그래프 표현, GNN 기반 비용 학습, 그리고 병원 간 연합 집계 기능을 결합한 연합 그래프 기반 쿼리 최적화 모델인 FedGQO를 제안한다. FedGQO는 원시 EHR 데이터나 상세한 로컬 쿼리 로그를 공유하지 않고도 협업 학습을 가능하게 하며, 비독립 동일분포(non-IID) 워크로드 하에서도 견고함을 유지한다. 실험 결과, FedGQO는 MSE 8.98, 평균 q-오차 1.21, 평균 실행 시간 8.40을 달성하여, 중앙 집중식 상한선에 근접하면서도 기본 최적화기 및 Local-GNN을 능가하는 성능을 보였다. 이러한 결과는 FedGQO가 병원 간 전자의무기록 쿼리 최적화를 위한 효과적이고 개인정보를 보호하는 솔루션임을 입증한다.

Cross-hospital EHR systems face privacy and workload diversity challenges that limit centralized learned optimization. We aim to build a federated query optimization framework that improves cost prediction and query execution efficiency while preserving institutional autonomy. We propose FedGQO, a federated graph-based query optimization model that combines plan-graph representation, GNN-based cost learning, and federated aggregation across hospitals. FedGQO enables collaborative training without sharing raw EHR data or detailed local query logs, while remaining robust under non-IID workloads. In experiments, it achieved an MSE of 8.98, a mean q-error of 1.21, and an average runtime of 8.40, outperforming the native optimizer and Local-GNN while remaining close to the centralized upper bound. These results show that FedGQO is an effective and privacy-preserving solution for cross-hospital EHR query optimization.

Artificial intelligence (AI) is a significant tool in modern military operations in that it helps to analyze a large volume of strategic, tactical, and operational data. On the other hand, current large language models (LLMs) like GPT-4 or Falcon have difficulty resolving problems in defense-specific contexts because of issues related to security, data confidentiality, and the lack of explainability. This document presents MilGPT, a secure and explainable LLM structure that aims at solving military problems only. To the model, fine-tuned open-source architectures with domain-specific defense datasets are integrated to elevate intelligence synthesis, decision-making, and threat prediction. On the benchmark, performance evaluation tasks show that MilGPT accounts for a 27% increase in contextual accuracy, an 18% reduction in hallucination rate, and an 33% improvement in explainability as measured by gradient-based feature attribution. In the proposed framework, military intelligence systems are not only secured but also made adaptive and humaninterpretable, thus, setting up a basis for the coming generation of AI models capable of defense-grade tasks.

Advanced Persistent Threats (APTs) represent a major headache for Security Operations Centers (SOCs) of the 21st century. Apart from being able to withstand constant monitoring, detection, and response, APTs are also extremely sophisticated and stealthy in nature. Centralized Intrusion Detection Systems (IDS), which are of a traditional nature, are usually not capable of providing adaptive, privacy-preserving, and collaborative detection functionalities across distributed networks. In this paper, a Federated Intrusion Detection Framework (FIDF), which uses Federated Learning (FL) to allow multiple SOC nodes to jointly train a smart detection model without the need to share raw data, is introduced. Local IDS agents, a central SOC aggregator, and a secure threat intelligence exchange mechanism are components of the system. The experimental performance is successful in showing that the detection accuracy is improved, the false positives are reduced, and the response time is enhanced for APT defense. The presented framework is a step toward federated solutions for the creation of a cybersecurity ecosystem that is scalable, privacy-aware, and resilient and is suitable for national defense infrastructures.

We investigate a data quality-aware dynamic client selection problem for multiple federated learning (FL) services in a wireless network, where each client offers dynamic datasets for the simultaneous training of multiple FL services, and each FL service demander has to pay for the clients under constrained monetary budgets. The problem is formalized as a non-cooperative Markov game over the training rounds. A multi-agent hybrid deep reinforcement learning-based algorithm is proposed to optimize the joint client selection and payment actions, while avoiding action conflicts. Simulation results indicate that our proposed algorithm can significantly improve training performance.

Deep reinforcement learning can effectively address resource allocation in wireless networks. However, its learning speed may be slower in more complex networks and a new policy should be learned for a newly-arrived system due to a lack of network adaptability. To address these issues, we propose a federated learning framework for resource allocation in wireless networks with multiple systems. It accelerates the learning speed by aggregating the policy at each system into a central policy and ensures network adaptability by using the central policy. Through experiments, we demonstrate that our proposed framework achieves both learning acceleration and network adaptability.

There is a great demand for an efficient security framework which can secure IoT systems from potential adversarial attacks. However, it is challenging to design a suitable security model for IoT considering the dynamic and distributed nature of IoT. This motivates the researchers to focus more on investigating the role of machine learning (ML) in the designing of security models. A brief analysis of different ML algorithms for IoT security is discussed along with the advantages and limitations of ML algorithms. Existing studies state that ML algorithms suffer from the problem of high computational overhead and risk of privacy leakage. In this context, this review focuses on the implementation of federated learning (FL) and deep learning (DL) algorithms for IoT security. Unlike conventional ML techniques, FL models can maintain the privacy of data while sharing information with other systems. The study suggests that FL can overcome the drawbacks of conventional ML techniques in terms of maintaining the privacy of data while sharing information with other systems. The study discusses different models, overview, comparisons, and summarization of FL and DL-based techniques for IoT security.

Federated Learning (FL) facilitates the design and training of deep learning (DL) models across multiple institutions and departments. For users with limited private data, FL can enhance the performance of local models by leveraging collective data from various sources. However, the devices participating in FL often differ in terms of network conditions, hardware and power capabilities. This results in heterogeneity across the computing, storage, and communication capabilities of the devices involved. Consequently, FL faces significant challenges due to device heterogeneity. Ignoring the inherent differences among clients when implementing FL protocols can significantly hinder both the efficiency and effectiveness of the training process. With the wide adaptability of FL for training DL models in resourceconstraint environments, the challenges caused by device heterogeneity have become increasingly apparent, highlighting the need for innovative strategies to minimize its impact on model aggregation. Therefore, this paper aims to review some of the most practical and effective existing proposed solutions. Future research can benefit from these solutions, as they can reduce the impact of device heterogeneity in FL.

This paper examines the integration of Federated Learning (FL), TinyML, and IoT in resource-constrained edge devices, highlighting key challenges and opportunities. It reviews FL and TinyML frameworks with a focus on communication, privacy, accuracy, efficiency, and memory constraints. We propose a novel FL-IoT framework that combines over-the-air (OTA) AI model updates, LoRa-based distributed communication, and lossless data compression techniques such as Run-Length Encoding (RLE), Huffman coding, and LZW to reduce transmission cost, optimize local processing, and maintain data privacy. The framework features Raspberry Pi-based aggregation nodes and microcontroller-based IoT clients, enabling scalable, low-power learning across heterogeneous devices. Evaluation includes memory usage, communication cost, energy consumption, and accuracy trade-offs across multiple FL scenarios. Results show improved scalability and significant power savings compared to baseline FL setups. The proposed framework is particularly impactful in applications such as smart agriculture, healthcare, and smart cities. Future directions for real-time, privacy-preserving edge intelligence are discussed.

As machine learning technology advances, data security becomes increasingly important. In this study, we propose an intrusion detection mechanism based on federated learning (FL) that updates only the learning weights to minimize the risk of information leakage. Considering the limited resources of industrial Internet of Things (IIoT) nodes, we propose a learning method based on data pruning. The proposed FL-based intrusion detection model was found to be more secure than the centralized model in terms of the data leakage rate. Data pruning technology reduced the memory usage by 1.4 times while maintaining 97.7 % accuracy. The proposed method detects attacks in industrial sites where large-scale IIoT nodes are installed efficiently, and protects industrial secrets and personal information effectively.

In recent years, advances in communication technology, hardware, and deep learning have led to significant advancements in Unmanned Aerial Vehicles (UAVs). However, applying federated learning in UAV environments is challenging due to network instability and dependency on central server. In this paper, the Federated Learning with Minimum Spanning Tree (MST)-based Weighted Tree Aggregation (FedWT) is proposed to address transmission failure, delayed model updates, and single point of failure problems. FedWT uses MST to minimize model exchange during local aggregation, and addresses data heterogeneity through dynamic weighted averaging. It includes a decentralized federated learning method for UAVs, model sharing path scheduling to reduce communication overhead, and a flexible weight-based aggregation approach to handle data heterogeneity. Simulation results demonstrate the superior performance and communication efficiency of FedWT compared to traditional federated learning methods. In particular, FedWT achieves up to 2% higher prediction accuracy and reduces communication traffic by up to 84.98% in highly heterogeneous data scenarios. Under different network topologies, FedWT consistently outperformed other federated learning methods in terms of learning accuracy and loss reduction. Future work will optimize the dynamic weight adjustment and validate the robustness of FedWT under various scenarios.