5G network slicing offers the potential to tailor network resources to diverse service requirements, but efficient and personalized resource allocation remains a significant challenge. Traditional centralized approaches struggle with scalability, privacy concerns, and the dynamic nature of user demands. This paper proposes a novel Federated Deep Reinforcement Learning (FDRL) framework for personalized resource allocation in 5G network slicing. The framework leverages federated learning to train a global deep reinforcement learning agent collaboratively across multiple edge servers, without sharing raw user data. Each edge server acts as a local agent, learning optimal resource allocation policies based on its local user data and contributing to the global model update. The proposed FDRL framework is designed to address the limitations of centralized approaches by enabling personalized resource allocation while preserving user privacy and enhancing scalability. We evaluate the performance of the FDRL framework through extensive simulations, demonstrating its superiority over centralized and non-federated DRL approaches in terms of resource utilization, service satisfaction, and privacy preservation. Furthermore, we analyze the impact of key parameters, such as the number of federated clients and the degree of data heterogeneity, on the performance of the FDRL framework.
Multivariate time series anomaly detection is a crucial task in various domains, including industrial monitoring, cybersecurity, and healthcare. Traditional methods often struggle to capture complex temporal dependencies and inter-variable correlations within high-dimensional data. This paper proposes a novel hybrid deep learning architecture that combines Long Short-Term Memory (LSTM) networks, Convolutional Neural Networks (CNNs), and attention mechanisms for enhanced anomaly detection in multivariate time series data. The LSTM network captures temporal dependencies, while the CNN extracts spatial features from the time series. Attention mechanisms are incorporated to focus on the most relevant features and time steps for anomaly detection. Furthermore, we introduce a feature engineering approach to derive meaningful features from the raw time series data, which are then fed into the deep learning model. The proposed approach is evaluated on several benchmark datasets and compared to state-of-the-art anomaly detection methods. The results demonstrate that our hybrid architecture with attention mechanisms and feature engineering significantly improves anomaly detection performance in terms of precision, recall, and F1-score.
Accurate time series forecasting is crucial for optimizing operations and decision-making in dynamic industrial environments. This paper proposes a novel hybrid deep learning framework that integrates Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolutional Neural Network (CNN) architectures to capture both temporal dependencies and local patterns within time series data. The framework is designed to adapt to the non-stationary nature of industrial processes, incorporating mechanisms for anomaly detection and robust performance in the presence of noise and outliers. We evaluate the performance of the proposed framework on real-world industrial datasets, demonstrating its superior accuracy and robustness compared to traditional time series forecasting methods and individual deep learning models. Furthermore, we analyze the impact of different hyperparameters and architectural configurations on the forecasting performance, providing insights into the optimal design of hybrid deep learning models for industrial time series data. The results highlight the potential of the proposed framework for predictive maintenance, resource optimization, and improved operational efficiency in dynamic industrial settings.
Industrial Control Systems (ICS) are increasingly vulnerable to sophisticated cyberattacks, posing significant threats to critical infrastructure. Traditional security measures often prove inadequate against advanced persistent threats (APTs) and zero-day exploits. This paper proposes a novel hybrid deep learning architecture for enhanced intrusion detection in ICS environments. The architecture leverages feature fusion techniques to combine diverse network traffic characteristics and employs an attention mechanism to selectively focus on the most relevant features for accurate anomaly detection. The proposed model integrates Convolutional Neural Networks (CNNs) for local pattern extraction and Recurrent Neural Networks (RNNs), specifically Gated Recurrent Units (GRUs), for capturing temporal dependencies in network traffic. Experimental results on a benchmark ICS dataset demonstrate the superior performance of the proposed hybrid model compared to state-of-the-art intrusion detection systems, achieving higher detection accuracy and lower false positive rates. The improved performance highlights the effectiveness of the feature fusion and attention mechanism in enhancing the model's ability to identify subtle and complex attack patterns in ICS networks.
Industrial Control Systems (ICS) are increasingly vulnerable to cyberattacks, necessitating robust anomaly detection mechanisms. This paper proposes an enhanced anomaly detection framework for ICS environments leveraging hybrid deep learning architectures and federated learning. We combine Long Short-Term Memory (LSTM) networks for temporal sequence analysis with Variational Autoencoders (VAEs) for feature reconstruction and outlier identification. To address data privacy concerns and the decentralized nature of ICS deployments, we implement a federated learning approach, enabling model training across multiple sites without sharing raw data. The proposed framework is evaluated on a publicly available ICS dataset, demonstrating superior performance compared to standalone deep learning models and traditional anomaly detection techniques. The results showcase improved accuracy, precision, and recall in identifying various attack types while preserving data privacy. The framework offers a practical and effective solution for enhancing the cybersecurity posture of modern ICS environments.