In today’s world, the volume of cyber attacks grows every year. These attacks can cause many people or companies high financial losses or loss of private data. One of the most common types of attack on the Internet is a DoS (denial-of-service) attack, which, despite its simplicity, can cause catastrophic consequences. A slow DoS attack attempts to make the Internet service unavailable to users. Due to the small data flows, these attacks are very similar to legitimate users with a slow Internet connection. Accurate detection of these attacks is one of the biggest challenges in cybersecurity. In this paper, we implemented our proposal of eleven major and most dangerous slow DoS attacks and introduced an advanced attack generator for testing vulnerabilities of protocols, servers, and services. The main motivation for this research was the absence of a similarly comprehensive generator for testing slow DoS vulnerabilities in network systems. We built an experimental environment for testing our generator, and then we performed a security analysis of the five most used web servers. Based on the discovered vulnerabilities, we also discuss preventive and detection techniques to mitigate the attacks. In future research, our generator can be used for testing slow DoS security vulnerabilities and increasing the level of cyber security of various network systems.

In this paper, possibilities for network traffic protection in future hybrid passive optical networks are presented, and reasons for realizing and utilizing advanced network traffic protection schemes for various network traffic classes in these networks are analyzed. Next, principles of the Prediction-based Fair Wavelength and Bandwidth Allocation (PFWBA) algorithm are introduced in detail, focusing on the Prediction-based Fair Excessive Bandwidth Reallocation (PFEBR) algorithm with the Early Dynamic Bandwidth Allocation (E-DBA) mechanism and subsequent Dynamic Wavelength Allocation (DWA) scheme. For analyzing various wavelength allocation possibilities in Hybrid Passive Optical Networks (HPON) networks, a simulation program with the enhancement of the PFWBA algorithm is realized. Finally, a comparison of different methods of the wavelength allocation in conjunction with specific network traffic classes is executed for future HPON networks with considered protection schemes. Subsequently, three methods are presented from the viewpoint of HPON network traffic protection possibilities, including a new approach for the wavelength allocation based on network traffic protection assumptions.

The interconnection of Operational Technology (OT) and Information Technology (IT) has created new opportunities for remote management, data storage in the cloud, real-time data transfer over long distances, or integration between different OT and IT networks. OT networks require increased attention due to the convergence of IT and OT, mainly due to the increased risk of cyber-attacks targeting these networks. This paper focuses on the analysis of different methods and data processing for protocol recognition and traffic classification in the context of OT specifics. Therefore, this paper summarizes the methods used to classify network traffic, analyzes the methods used to recognize and identify the protocol used in the industrial network, and describes machine learning methods to recognize industrial protocols. The output of this work is a comparative analysis of approaches specifically for protocol recognition and traffic classification in OT networks. In addition, publicly available datasets are compared in relation to their applicability for industrial protocol recognition. Research challenges are also identified, highlighting the lack of relevant datasets and defining directions for further research in the area of protocol recognition and classification in OT environments.

This paper presents a comprehensive investigation of machine learning-based intrusion detection methods to reveal cyber attacks in railway axle counting networks. In contrast to the state-of-the-art works, our experimental results are validated with testbed-based real-world axle counting components. Furthermore, we aimed to detect targeted attacks on axle counting systems, which have higher impacts than conventional network attacks. We present a comprehensive investigation of machine learning-based intrusion detection methods to reveal cyber attacks in railway axle counting networks. According to our findings, the proposed machine learning-based models were able to categorize six different network states (normal and under attack). The overall accuracy of the initial models was ca. 70–100% for the test data set in laboratory conditions. In operational conditions, the accuracy decreased to under 50%. To increase the accuracy, we introduce a novel input data-preprocessing method with the denoted gamma parameter. This increased the accuracy of the deep neural network model to 69.52% for six labels, 85.11% for five labels, and 92.02% for two labels. The gamma parameter also removed the dependence on the time series, enabled relevant classification of data in the real network, and increased the accuracy of the model in real operations. This parameter is influenced by simulated attacks and, thus, allows the classification of traffic into specified classes.

The creation of suitable and efficient tools for anomaly detection constitutes a crucial aspect of security, applicable not only to industrial networks but also to cyber-physical systems. This article elucidates a framework designed to automate the selection of an optimal deep neural network architecture, thereby expediting the creation and implementation of neural network-based tools. The framework presented here enables a rapid design of an Artificial Neural Network structure without necessitating user intervention. Its efficacy has been showcased through experimentation with the publicly accessible HAI dataset, yielding an accuracy of approximately 0.94 after 10 epochs. Subsequently, a second scenario was performed where a total of 5456 models were generated and trained, with an average time of approximately 9.95 seconds per model.