Network Anomaly Detection Using Wavelet Neural Networks Trained with Quantum-Behaved Particle Swarm Optimization

In network anomaly detection, to enhance the detection rate of anomalous states and minimize the misjudgment rate of normal states, this paper introduces a novel approach that utilizes Quantum-behaved Particle Swarm Optimization (QPSO) to train Wavelet Neural Networks (WNN) for anomaly detection. The parameter combination within the WNN is represented as a particle in the optimization algorithm, enabling a global search for the parameter vector with the best fitness value. The QPSO algorithm updates particle positions based on quantum behavior, such as wave function collapse and potential well attraction, which improves search efficiency and avoids local optima. Additionally, to further boost the accuracy of network anomaly detection, we incorporate feature selection techniques to identify the most representative feature subsets from raw data. This involves methods like correlation analysis or wrapper approaches to reduce dimensionality and enhance model performance. By integrating these methodologies, our research offers a new and effective solution for the field of network anomaly detection, with potential code implementation including QPSO-based parameter optimization loops and wavelet activation functions in neural network layers.