Investigation of computer systems to detect intrusions and network anomalies

Abstract

The article describes models of intrusion and network anomaly detection systems with quantum autocoding in computer systems. The paper proposes innovative methods for researching intrusion and network anomaly detection systems with quantum autocoding in computer systems that can provide fast response and a high level of adaptability. The paper proposes a quantum QAE (Quantised Autoencoder) model is used in intrusion detection systems to identify anomalies. This model is an optimization approach based on autoencoders, which integrates techniques such as cut-off, clustering, and integer quantisation.

Relevance. The significance of this work lies in the ability to investigate intrusion and network anomaly detection systems utilizing quantum autoencoding in information and communication systems. The study focuses on creating a method for detecting anomalous attacks in IoT network traffic, as identifying anomalies requires detailed monitoring of various network activities. Moreover, the network traffic of each IoT device is distinct. Consequently, the study applies an autoencoder algorithm for anomaly detection, using benign network traffic for model training, with the assumption that any anomalous traffic would lead to an anomaly reconstruction (AR) error.

Research methods. methods for studying intrusion detection systems and network anomalies with quantum autocoding in information and communication systems are probabilistic, verification modelling, and the use of cloud computing, which provide flexibility, scalability and resources for building effective computer attack detection systems.

The results. A real-time IoT dataset was created for both normal and attack traffic. During the training phase, the autoencoder model is trained on normal traffic. The same model is then used to reconstruct anomalous traffic, with the expectation that the reconstruction error (RE) for anomalies will be significant, aiding in the detection of attacks. Additionally, the performance of the autoencoder model was evaluated using metrics such as precision, accuracy, recall, and through a comprehensive experimental study.

Conclusions. The results show that there is a trade-off between the autoencoder and the QAE-u8 model in terms of accuracy and processor evaluation parameters such as memory and CPU. Thus, we conclude that there is a trade-off between the autoencoder and the QAE-u8 model in terms of accuracy and processor evaluation parameters such as memory and CPU. In future research, we will focus on other IoT device vulnerabilities to develop a more secure IoT infrastructure.

The scientific novelty of this work is the development of strategies and techniques for identifying anomalous attacks in IoT network traffic.