基于量子神经网络的网络安全态势感知

Network security situational awareness based on quantum neural network

网络安全态势感知的预测方法大都预测准确度较低,收敛速度较慢;采用传统的神经网络模型,其模型设置参数多,泛化性能不高,无法进行有效预测。基于上述问题,提出了一种利用量子粒子群优化算法(Quantum Particle Swarm Optimization, QPSO)来优化变分量子神经网络(Variational Quantum Neural Network, VQNN)的预测模型,其训练过程是将数据样本经过预处理后转化为量子态进行输入,并通过隐藏层的酉变换反复调整权重参数θ,然后利用改进惯性权重因子θ1对权重参数θ进行优化来获取网络的输出,将优化后的VQNN模型用于预测网络安全态势值。实验结果表明,VQNN模型的精确度比卷积神经网络模型高0.47%,且收敛速度快,将QPSO-VQNN算法与对比预测方法进行实验对比,验证了所提算法在网络安全态势感知中的可行性以及预测结果的准确性。

Most of the prediction methods for network security situation awareness have the following problems:The prediction accuracy is low,and the convergence speed is slow;The generalization performance of the traditional neural network,which has many model parameters,is poor and effective predicting is impossible.Based on the above-mentioned problems,a prediction model using a quantum particle swarm optimization(QPSO)algorithm is proposed to optimize the variational quantum neural network(VQNN).After preprocessing,the data sample is converted into a quantum state input,and the weighted parameter is adjusted periodically using the hidden layer's unitary transformation,and then improved inertia weight factor is used to optimize weighted parameter to obtain the output of the network,and the optimized VQNN model is used to predict network security situation value.The experimental results show that the accuracy of VQNN model used in this paper is 0.47%higher than convolution neural network model,and the convergence speed is faster.By comparing the proposed QPSO-VQNN algorithm with comparative prediction methods,the feasibility of the proposed algorithm in network security situational awareness and the accuracy of prediction results are verified.