In recent years, due to the rapid growth in network technology, numerous types of intrusions have been uncovered that differ from the existing ones, and the conventional firewalls with specific rule sets and policies are incapable of identifying those intrusions in real-time. Therefore, that demands the requirement of a real-time intrusion detection system (RT-IDS). The ultimate purpose of this research is to construct an RT-IDS capable of identifying intrusions by analysing the inbound and outbound network data in real-time. The proposed system consists of a deep neural network (DNN) trained using 28 features of the NSL-KDD dataset. In addition, it contains the machine learning (ML) pipeline with sequential components for categorical data encoding and feature scaling, which is used before transmitting the real-time data to the trained DNN model to make predictions. Moreover, a real-time feature extractor, which is a C++ program that sniffs data from the real-time network traffic and derives relevant data related to the features of the NSL-KDD dataset using the sniffed data, is deployed between the gateway router and the local area network (LAN). Together with the trained DNN model, the ML pipeline is hosted in a server that can be accessed via a representational state transfer application programming interface (REST API). The DNN has revealed outstanding testing performance results achieving 81%, 96%, 70% and 81% for accuracy, precision, recall and f1-score accordingly. This research comprises a comprehensive technical explanation concerning the implementation and functionality of the complete system. Moreover, leveraging the extensive explanations provided in this paper, advanced IDSs capable of identifying modern intrusions can be constructed.

Funded by European Framework-7 (FP7), the CockpicCI project aims at developing intelligent risk detection, analysis and protection techniques for Critical Infrastructures (CI). In this paper, we describes our recent research on automated anomaly detection from central Supervisory Control and Data Acquisition (SCADA) systems and their related commands/measurements in the SCADA-field equipment communications. The work exploits the concept of one-class SVM (Support Vector Machines) and adaptively controls its decision parameter to detect unusual patterns from inputs and generate alarms for on-site engineers to further investigate. Experiments on simulation data sets from telecommunication networks illustrate that the proposed algorithm achieves high detection rates, providing excellent potential for further research and development towards practical tools for protection of SCADA systems.

Accurate classification of cancer images plays a crucial role in diagnosis and treatment planning. Deep learning (DL) models have shown promise in achieving high accuracy, but their performance can be influenced by variations in Hematoxylin and Eosin (H&E) staining techniques. In this study, we investigate the impact of H&E stain normalization on the performance of DL models in cancer image classification. We evaluate the performance of VGG19, VGG16, ResNet50, MobileNet, Xception, and InceptionV3 on a dataset of H&E-stained cancer images. Our findings reveal that while VGG16 exhibits strong performance, VGG19 and ResNet50 demonstrate limitations in this context. Notably, stain normalization techniques significantly improve the performance of less complex models such as MobileNet and Xception. These models emerge as competitive alternatives with lower computational complexity and resource requirements and high computational efficiency. The results highlight the importance of optimizing less complex models through stain normalization to achieve accurate and reliable cancer image classification. This research holds tremendous potential for advancing the development of computationally efficient cancer classification systems, ultimately benefiting cancer diagnosis and treatment.

SCADA (Supervisory Control And Data Acquisition) systems have always been susceptible to cyber-attacks. Different types of cyber-attacks could occur depending on the architecture and configurations used in the SCADA system. To protect cyber infrastructure from above attacks a growing collaborative effort between cyber security professionals and researchers from private and academia has involved in designing variety of intelligent intrusion detection systems. This paper introduces a new European Framework-7 project CockpitCI and roles of intelligent machine learning methods to prevent SCADA systems from cyber-attacks.

Restoring strength, function, and mobility following an illness, accident, or surgery is the primary goal of upper arm rehabilitation. Exoskeletons offer adaptable support, enhancing patient engagement and accelerating recovery. This work proposes an adjustable, wearable robotic exoskeleton powered by electromyography (EMG) data for upper arm rehabilitation. Three activation levels—low, medium, and high—were applied to the EMG data to forecast the Pulse Width Modulation (PWM) based on the range of motion (ROM) angle. Conventional machine learning (ML) models, including K-Nearest Neighbor Regression (K-NNR), Support Vector Regression (SVR), and Random Forest Regression (RFR), were compared with neural network approaches, including Gated Recurrent Units (GRUs) and Long Short-Term Memory (LSTM) to determine the best ML model for the ROM angle prediction. The LSTM model emerged as the best predictor with a high accuracy of 0.96. The system achieved 0.89 accuracy in exoskeleton control and 0.85 accuracy in signal categorization. Additionally, the proposed exoskeleton demonstrated a 0.97 performance in ROM correction compared to conventional methods (p = 0.097). These findings highlight the potential of EMG-based, LSTM-enabled exoskeleton systems to deliver accurate and adaptive upper arm rehabilitation, particularly for senior citizens, by providing personalized and effective support.

In recent years, various types of new intrusions that differ from existing ones have been identified. Moreover, due to the rapid evolution of cyberattacks, machine learning algorithms require updated datasets that comprise the most recent intrusions. The prime objective of this research is to discover the best machine learning algorithm for intrusion detection trained using the NSL-KDD and the UNSW-NB15 datasets and perform a comparative analysis between six machine learning algorithms classified as supervised, semi-supervised, and unsupervised learning. This study revealed that the performance of supervised and semi-supervised machine learning algorithms outperformed unsupervised machine learning algorithms for both datasets and concluded that Support Vector Machines (SVM) and Deep Neural Network (DNN) perform better for NSL-KDD and UNSW-NB15, respectively.