LLM-Based Edge Intelligence: A Comprehensive Survey on Architectures, Applications, Security and Trustworthiness

IEEE Open Journal of the Communications Society, Journal Year: 2024, Volume and Issue: 5, P. 5799 - 5856

Published: Jan. 1, 2024

Language: Английский

---

Advancements in securing federated learning with IDS: a comprehensive review of neural networks and feature engineering techniques for malicious client detection

Artificial Intelligence Review, Journal Year: 2025, Volume and Issue: 58(3)

Published: Jan. 13, 2025

Federated Learning (FL) is a technique that can learn global machine-learning model at central server by aggregating locally trained models. This distributed approach preserves the privacy of local However, FL systems are inherently vulnerable to significant security challenges such as cyber-attacks, handling non-independent and identically (non-IID) data, data concerns. systematic literature review addresses these issues examining advanced neural network models, feature engineering methods, privacy-preserving techniques within intrusion detection (IDS) for environments. These key elements improving systems. To best our knowledge, this among first comprehensively explore combined impacts technologies. We analyzed 88 studies published between 2021 October 2024. study offers valuable insights future research directions, including scaling in real-world environment.

Language: Английский

---

Intelligent deep federated learning model for enhancing security in internet of things enabled edge computing environment

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 3, 2025

In the present scenario, Internet of Things (IoT) and edge computing technologies have been developing rapidly, foremost to development new tasks in security privacy. Personal information privacy leakage become main concerns IoT surroundings. The promptly IoT-connected devices below an integrated Machine Learning (ML) method might threaten data confidentiality. standard centralized ML-assisted methods challenging because they require vast numbers a vital unit. Due rising distribution many systems linked devices, decentralized ML solutions required. Federated learning (FL) was proposed as optimal solution discover these issues. Still, heterogeneity environments poses essential task when executing FL. Therefore, this paper develops Intelligent Deep Model for Enhancing Security (IDFLM-ES) approach IoT-enabled edge-computing environment. presented IDFLM-ES aims identify unwanted intrusions certify safety To accomplish this, technique introduces federated hybrid deep belief network (FHDBN) model using FL on time series produced by devices. Besides, uses normalization golden jackal optimization (GJO) based feature selection pre-processing step. learns individual distributed representation over databases enhance convergence quick learning. Finally, dung beetle optimizer (DBO) is utilized choose effectual hyperparameter FHDBN model. simulation value methodology verified benchmark database. experimental validation portrayed superior accuracy 98.24% compared other models.

Language: Английский

---

Federated learning based reference evapotranspiration estimation for distributed crop fields

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(2), P. e0314921 - e0314921

Published: Feb. 5, 2025

Water resource management and sustainable agriculture rely heavily on accurate Reference Evapotranspiration (ET o ). Efforts have been made to simplify the ) estimation using machine learning models. The existing approaches are limited a single specific area. There is need for ET estimations of multiple locations with diverse weather conditions. study intends propose distinct conditions federated approach. Traditional centralized require aggregating all data in one place, which can be problematic due privacy concerns transfer limitations. However, trains models locally combines knowledge, resulting more generalized estimates across different regions. three geographical Pakistan, each conditions, selected implement proposed model from 2012 2022 locations. At location, named Random Forest Regressor (RFR), Support Vector (SVR), Decision Tree (DTR), evaluated local (ET) global model. feature importance-based analysis also performed assess impacts parameters performance at location. evaluation reveals that (RFR) based outperformed other coefficient determination (R 2 = 0.97%, Root Mean Squared Error (RMSE) 0.44, Absolute (MAE) 0.33 mm day −1 , Percentage (MAPE) 8.18%. yields against site. results suggest maximum temperature wind speed most influential factors predictions.

Language: Английский

---

Comparative Review of the Intrusion Detection Systems Based on Federated Learning: Advantages and Open Challenges

Algorithms, Journal Year: 2022, Volume and Issue: 15(7), P. 247 - 247

Published: July 15, 2022

In order to provide an accurate and timely response different types of the attacks, intrusion anomaly detection systems collect analyze a lot data that may include personal other sensitive data. These could be considered source privacy-aware risks. Application federated learning paradigm for training attack models significantly decrease such risks as generated locally are not transferred any party, is performed mainly on sources. Another benefit usage its ability support collaboration between entities share their dataset confidential or reasons. While this approach able overcome aforementioned challenges it rather new well-researched. The research questions appear while using implement analytical systems. paper, authors review existing solutions based learning, study advantages well open still facing them. paper analyzes architecture proposed approaches used model partition across clients. ends with discussion formulation challenges.

Language: Английский

---