Federated Learning for Privacy-Preserving Medical Data Analytics in Big Data

Published: May 9, 2024

Language: Английский

---

Privacy-Preserving Federated Learning-Based Intrusion Detection System for IoHT Devices

Electronics, Journal Year: 2024, Volume and Issue: 14(1), P. 67 - 67

Published: Dec. 27, 2024

In recent years, Internet of Healthcare Things (IoHT) devices have attracted significant attention from computer scientists, healthcare professionals, and patients. These enable patients, especially in areas without access to hospitals, easily record transmit their health data medical staff via the Internet. However, analysis sensitive information necessitates a secure environment safeguard patient privacy. Given sensitivity data, ensuring security privacy is crucial this sector. Federated learning (FL) provides solution by enabling collaborative model training sharing with third parties. Despite FL addressing some concerns, IoHT remains an area needing further development. paper, we propose privacy-preserving federated framework enhance data. Our approach integrates ϵ-differential design effective intrusion detection system (IDS) for identifying cyberattacks on network traffic devices. our FL-based framework, SECIoHT-FL, employ deep neural (DNN) including convolutional (CNN) models. We assess performance SECIoHT-FL using metrics such as accuracy, precision, recall, F1-score, budget (ϵ). The results confirm efficacy efficiency framework. For instance, proposed CNN within achieved accuracy 95.48% (ϵ) 0.34 when detecting attacks one datasets used experiments. To facilitate understanding models reproduction experiments, provide explainability SHAP share source code publicly free open-source software.

Language: Английский

---

Healthcare 5.0 Fundamentals

Advances in healthcare information systems and administration book series, Journal Year: 2023, Volume and Issue: unknown, P. 1 - 25

Published: Dec. 18, 2023

Healthcare 5.0 signifies a radical paradigm shift in the healthcare sector an era of technology that is advancing at exponential rate. In this chapter, author goes into fundamental ideas and real-world uses support revolution. The historical view presented chapter shows how concepts have changed through time, from earlier iterations to current 5.0. It highlights crucial part has played influencing new healthcare.

Language: Английский

---

MULTI-KEY FULLY HOMOMORPHIC ENCRYPTION FOR PRIVACY-PRESERVATION WITHIN FEDERATED LEARNING ENVIRONMENTS

EDPACS, Journal Year: 2023, Volume and Issue: 68(6), P. 25 - 34

Published: Dec. 2, 2023

Despite the need for data from multiple sources in machine learning, privacy constraints limit sharing. Federated Learning (FL) addresses this by allowing clients to share locally trained model parameters without disclosing sensitive data, however, recent research highlights leakage risks. This paper investigates multi-key fully homomorphic encryption, specifically MK-CKKS, enhance FL. The study demonstrates MK-CKKS’s effectiveness protecting transmission and preventing external access private information. Nonetheless, precautions are needed during decryption, as vulnerabilities may allow aggregator server adversaries infer personal shared partial descriptions, impacting client’s security.

Language: Английский

---

Modelagem das Áreas de Risco de Sistemas de Detecção de Intrusão para Cálculo de Métricas de Privacidade

Published: Sept. 18, 2023

Embora regulamentações recentes exijam que desenvolvedores de software passem a se preocupar forma severa com privacidade, recomendações relacionadas isso existem há anos. Apesar do entendimento, relativamente antigo, sistemas computacionais precisam garantir privacidade usuário, tem sido difícil encontrar trabalhos avaliem as em existentes, principalmente porque métricas variam o domínio das aplicações. Este artigo apresenta resultados preliminares decorrentes da modelagem áreas risco visando medição um IDS baseado aprendizado máquina. É mostrado foi possível adaptar princípios literatura para nosso domínio.

---

IoT-Cloud Integration with Reinforcement Learning for Elderly Fall Detection

Published: Dec. 29, 2023

This project greatly contributes to the integration of IOT systems for actual development fall detection mechanisms with advanced RL algorithms. Hence system mainly prefers healthcare elderly individuals. The individual responses are evaluated according major A brief introduction is given here a overview IOT-cloud formation and its theoretical frameworks related maximum usage methodology part shows philosophy, approach, design detection. thematic analysis provided make descriptive result algorithms an efficient mechanism alarm structures.

Language: Английский

---

Enhancing Security and Privacy in Cloud – Based Healthcare Data Through Machine Learning

Published: Dec. 29, 2023

It is becoming more and important for healthcare providers to protect the integrity security of sensitive medical data as they use cloud computing processing storage. This work explores field machine learning algorithms that are secure privacy-preserving when applied information in environments. We investigate sophisticated cryptography, federated learning, differentiating privacy techniques using an interpretive philosophy a method based on deduction. Our results highlight computational expense associated with cryptographic protocols, while also revealing their nuanced performance potential enabling calculations. Federated shown be effective collaborative model training, providing workable approach analysis over-dispersed datasets. Differential systems require careful parameter calibration because demonstrate delicate balance between value preservation.

Language: Английский

---

Federated Learning for Privacy-Preserving Healthcare Data Analysis in the Age of Cybersecurity Threats

Published: Dec. 29, 2023

This examination explores joined picking up gathering appraisals, unequivocally United Averaging (FedAvg), Weighted Consolidated (FedAvg-W), Bound together Learning with Adaptable Rate (FedAdapt), and Secure Combination for Brought (SecAgg), inside the space of assertion saving clinical benefits data assessment. The reason organized assessments was to assess their performance in terms accuracy, evidence coverage communication speed. article provides a comparative evaluation help practitioners select most appropriate algorithm reasoning applications. results show that FedAvg-W achieves much higher accuracy than other algorithms especially when used locations varying attributes implying it can adapt changes. In relation this, method called FedAdapt mixes quickly while maintaining high by way dynamically changing learning rate limits respect particular instances distribution information. A secure aggregation framework based on homomorphic encryption guarantees exact compliance. review subtle experiences into space-related works, such as health informatics federated learning. On one hand, SecAgg fulfills basic requirement ensuring preserving medical side, FedAdapt's flexibility concerns anticipated scalability

Language: Английский

---