Research On Detection And Evaluation Methods Of Internal User Anomaly Behavior Based On Quantum Machine Learning

In the field of network security,malicious behavior from internal users poses a serious threat to systems and networks.In recent years,network security incidents caused by internal user misconduct have continued to occur,and the damage caused by such threats is even more severe than that caused by external attacks.These internal users typically have important privileges,such as access to internal systems,knowledge of system vulnerabilities,and confidential information.With the continuous development of technology,these malicious behaviors by internal users have become increasingly covert and diverse.Therefore,detecting abnormal behaviors from internal users has become a research hotspot in the current field of network security.Quantum computing takes advantage of properties such as superposition and parallelism of quantum states to exhibit astonishing capabilities in processing high-dimensional data.Currently,quantum computing is in the era of Noisy Intermediate-Scale Quantum(NISQ),which explores the potential applications of quantum computing and is a hot research direction.In the classical computing field,there have been some research achievements in anomaly detection.However,research on anomaly detection in the field of quantum computing is still lacking.In this thesis,we explore anomaly detection using quantum computing,design relevant quantum algorithms,and conduct simulation experiments.Using quantum computing to explore internal user anomaly detection and evaluation provides a new application scenario for quantum computing,while also providing new ideas for anomaly detection and evaluation.The specific work of this thesis is as follows:(1)Design a quantum convolutional neural network algorithm(QCNN-BD)for abnormal behavior detection by internal users.QCNN-BD is composed of a quantum convolutional neural network(QCNN),which is a neural network model that applies quantum computing and combines the advantages of quantum computing and deep learning.Its purpose is to solve problems in traditional deep learning,such as limitations in processing low-dimensional data or being easily trapped in local optimal solutions.The basic structure of QCNN is similar to that of traditional convolutional neural networks(CNNs),but considering the current NISQ,the QCNN designed in this work is implemented by combining quantum convolutional kernels with classical neural networks.Using the Penny Lane framework,we detect and analyze different user behaviors.The experimental results show that the designed QCNN-BD has higher computational efficiency and stronger processing capabilities in internal anomaly detection compared to traditional convolutional neural networks.(2)A prominent problem in abnormal behavior detection by internal users is the imbalance in the number of positive and negative samples.To address this issue,a quantum generative adversarial network(QGAN-BDE)was designed for internal user behavior detection and evaluation.The QGAN-BDE algorithm consists of a quantum generative adversarial network(QGAN)for generating negative samples,and a classical neural network for detection and evaluation.QGAN adopts a quantum-classical hybrid structure,namely,a quantum generator(G_Q)and a classical discriminator(_CD).G_Q is implemented by a parameterized quantum circuit(PQC),and _CD is a classical fully-connected neural network.The negative samples are generated by G_Q,and the adversarial game between G_Q and _CD is used to improve the sample generation ability of G_Q,thereby addressing the problem of the imbalance in the number of positive and negative samples.Through simulation testing and analysis,the QGAN-BDE algorithm can effectively detect and evaluate abnormal behavior by internal users.